Search for New Physics with b sll decays @ LHCb Simone Bifani University of Birmingham (UK) On behalf of the LHCb Collaboration CERN Seminar, 18th April 17
A Forward Spectrometer Optimized for beauty and charm physics at large pseudorapidity (<h<5)» Trigger: >95% (6-7%) efficient for muons (electrons)» Tracking: sp/p.4%.6% (p from 5 to 1 GeV), sip < µm» Calorimeter: se/e ~ 1% / E 1%» PID: ~97% µ,e ID for 1 3% p µ,e misid Simone Bifani CERN Seminar
Datasets Analysis presented today based on the full Run 1 dataset Dimuons per GeV/c 11 1 1 1 1 1 1 1 1 9 8 7 6 5 4 11+1 data Single muon Charmonium Bottomonium Other triggers ρ/ω φ J/ψ ψ(s) Υ(nS) 9 9.5 1 1.5 11 Dimuon mass [GeV/c ] 1 3 1 1 1 1. 1 3 4 5 1 1 LHCB-CONF-16-5 Dimuon mass [GeV/c ] Dimuons per GeV/c 3 1 6 1 4 Υ (1S) Z Υ (S) Υ (3S) Due to luminosity levelling, same running conditions throughout fills 3
Why Rare b Decays? b sll decays proceed via FCNC transitions that only occur at loop order (or beyond) in the SM SM b B d + W t, c, u γ, Z s l + l K d * New particles can for example contribute to loop or tree level diagrams by enhancing/suppressing decay rates, introducing new sources of CP violation or modifying the angular distribution of the final-state particles b B d Rare b decays place strong constraints on many NP models by probing energy scales higher than direct searches 4 Z NP? + W t, c, u b B d t, c, u l + + W - s l + l K d * W s l K d *
Shopping List Differential branching fractions of B K (*) µµ, B + K (*)+ µµ, B s fµµ, B + p + µµ and L b Lµµ» Presence of hadronic uncertainties in theory predictions Angular analyses of B K (*) µµ, B s fµµ, B K * ee and L b Lµµ» Define observables with smaller theory uncertainties Test of Lepton Flavour Universality in B + K + ll and B K * ll» Cancellation of hadronic uncertainties in theory predictions Different q regions probe different processes In the OPE framework the short-distance contribution is described by Wilson coefficients 5
Differential Branching Fractions Results consistently lower than SM predictions ] c 4 /GeV -8 [1 db/dq ] [c 4 /GeV db/dq 5 4 3 1.5 J/y y(s) 5 1 15 q [GeV JHEP 6 (14) 133.15 1.1 6 LCSR Lattice Data B K * µµ LHCb J/y 5 1 15 arxiv:166.4731 q [GeV /c 4 ] B + + B + K K µ + + µ µµ LHCb LHCb y(s) /c 4 ] c 4 ] - GeV -8 [1 φµµ)/dq db(bs 9 8 7 6 5 4 3 1 B s fµµ 3.3s form SM JHEP 9 (15) 179 ] -1 /c 4 ) (GeV -7 [1 µ) / dq Λ µ db(λ b 1.8 1.6 1.4 1. 1.8.6.4. SM prediction Data 5 1 15 q q LHCb SM pred. Data [GeV /c 4 ] LHCb 5 1 15 JHEP 6 (15) 115 L b Lµµ J/y J/y y(s) y(s) [GeV /c 4 ] 6
Angular Analyses First full angular analysis of B K * µµ: measured all CP-averaged angular terms and CP-asymmetries Can construct less form-factor dependent ratios of observables 5 P' 1.5 LHCb data Belle data ATLAS data CMS data SM from DHMV SM from ASZB.5 J/ψ(1S) ψ(s) 1 5 1 15 JHEP (16) 14 PRL 118 (17).8 and 3. s from SM ATLAS-CONF-17-3 CMS-PAS-BPH-15-8 q [GeV /c 4 ] 7
Once upon a time LHCb tested Lepton Universality using B + K + ll decays and observed a tension with the SM at.6s R K LHCb BaBar Belle LHCb 1.5 1 SM.5.6s form SM 5 1 15 PRL 113 (14) 15161 q PRD 86 (1) 31 PRL 13 (9) 17181 [GeV /c 4 ] Consistent with observed BR(B + K + µµ) if NP does not couple to electrons Observation of LFU violations would be a clear sign of NP 8
Global Fits Several attempts to interpret results by performing global fits to data 3 Branching Ratios. Angular Observables!Pi " All 1.5 1 1 1. NPNP Re( Re CC 1 1 ) CNP 1 ATLAS CMS LHCb BR only all!1.5. -1.5! - 1. flavio v..3!3!3!!1 1 CNP 9 arxiv:151.439 3 1.5. -3 1.5-1. -1.5. NP.5 11. 1.5 Re C9 NP Re( C9 ) arxiv:151.439 arxiv:173.9189 arxiv:1611.56 Take into account ~9 observables from different experiments, including B µµ and b sll transitions All global fits require an additional contribution with respect to the SM to accommodate the data, with a preference for NP in C9 at ~4s Or is this a problem with the understanding of QCD? (e.g. correctly estimating the contribution from charm loops?) Simone Bifani CERN Seminar 9
Today Test of LFU with B K * µµ and B K * ee, R K*º Two regions of q»low [.45-1.1] GeV /c 4»Central [1.1-6.] GeV /c 4 low-q central-q Measured relative to B K * J/y(ll) in order to reduce systematics K * reconstructed as K + p - within 1MeV from the K * (89) Blind analysis to avoid experimental biases Extremely challenging due to significant differences in the way µ and e interact with the detector»bremsstrahlung»trigger 1
Bremsstrahlung I Electrons emit a large amount of bremsstrahlung that results in degraded momentum and mass resolutions Two types of bremsstrahlung» Downstream of the magnet - photon energy in the same calorimeter cell as the electron - momentum correctly measured» Upstream of the magnet - photon energy in different calorimeter cells than electron - momentum evaluated after bremsstrahlung Upstream brem Air Downstream brem 11
Bremsstrahlung II A recovery procedure is in place to improve the momentum reconstruction Events are categorised depending on the number of recovered photon clusters Incomplete recovery due to 18 16 14 1 part-reco 1 8 6 central-q 4 low-q 45 5 14 y(s) 13 J/y 1 1 55 6 m(k π µ µ ) [MeV/c] 1 q [GeV/ c4] q [GeV/ c4]» Energy threshold of the bremsstrahlung photon (ET > 75 MeV)» Calorimeter acceptance» Presence of energy deposits mistaken as bremsstrahlung photons 18 16 y(s) 14 1 part-reco J/y 1 8 brem tail 6 central-q 4 low-q 45 5 55 13 1 1 6 1 m(k π ee) [MeV/c] Incomplete recovery causes the reconstructed B mass to shift towards lower values and events to migrate in and out of the q bins Simone Bifani CERN Seminar 1
Trigger Trigger system split in hardware (L) and software (HLT) stages Due to higher occupancy of the calorimeters compared to the muon stations, hardware thresholds on the electron E T are higher than on the muon p T (L Muon, p T >1.5,1.8 GeV) To partially mitigate this effect, 3 exclusive trigger categories are defined» L Electron: electron hardware trigger fired by clusters associated to at least one of the two electrons (E T >.5 GeV)» L Hadron: hadron hardware trigger fired by clusters associated to at least one of the K * decay products (E T > 3.5 GeV)» L TIS: any hardware trigger fired by particles in the event not associated to the signal candidate 13
Strategy R K*º determined as double ratio to reduce systematic effects Selection as similar as possible between µµ and ee» Pre-selection requirements on trigger and quality of the candidates» Cuts to remove the peaking backgrounds» Particle identification to further reduce the background» Multivariate classifier to reject the combinatorial background» Kinematic requirements to reduce the partially-reconstructed backgrounds» Multiple candidates randomly rejected (1-%) Efficiencies» Determined using simulation, but tuned using data 14
Corrections to Simulation Four-step procedure largely based on tag-and-probe technique 1. Particle identification» PID response of each particle species tuned using dedicated calibration samples. Generator» Event multiplicity and B kinematics matched to data using B K * J/y(µµ) decay 3. Trigger» Hardware and software trigger responses tuned using B K * J/y(ll) decays 4. Data/MC differences» Residual discrepancies in variables entering the MVA reduced using B K * J/y(ll) decays After tuning, very good data/mc agreement in all key observables 15
Fit Procedure µµ Fit signal MC to extract initial parameters Simultaneous fit to resonant and non-resonant data allowing (some) parameters to vary Signal» Hypatia [NIM A, 764, 15 (14)]» Free parameters mass shift and width scale Backgrounds» Combinatorial exponential» L b pk - J/y(µµ) simulation & data» B s K * J/y(µµ) same as signal but shifted by m Bs -m B B K * J/y only 16
Fit Results µµ Pulls Candidates per 1 MeV/c 9 8 7 6 5 4 3 1 5 5 Signal Combinatorial 5 54 56 58 m(kπ µµ) [MeV/c ] low-q central-q Pulls Candidates per 1 MeV/c 8 7 6 5 4 3 1 5 5 Signal Combinatorial 5 54 56 58 m(kπ µµ) [MeV/c ] Candidates per 1 MeV/c 4 1 3 1 1 1 5 Signal Combinatorial Λ b pkj/ ψ B s K * J/ ψ Pulls 5 5 54 56 58 m(kπ µµ) [MeV/c ] 17
Fit Procedure ee Fit signal MC to extract initial parameters Simultaneous fit to resonant and non-resonant data split in trigger categories allowing (some) parameters to vary (bremsstrahlung fractions fixed from MC) Signal» Crystal-Ball (Crystal-Ball and Gaussian)» Free parameters mass shift and width scale Backgrounds» Combinatorial exponential» L b pk - J/y(ee) simulation & data, constrained using muons» B s K * J/y(ee) same as signal but shifted by m Bs -m B, constrained using muons» B K * J/y Leakage simulation, yield constrained using data» Part-Reco simulation & data B K * J/y only B K * ee only 18
Part-Reco Background I Partially-reconstructed backgrounds arise from decays involving higher K resonances with one or more decay products in addition to a Kp pair that are not reconstructed Large variety of decays, most abundant due to B K 1 (17)ee and B K * (143)ee 19
Part-Reco Background II Modelled using two independent methods»create a K 1 +K cocktail from simulation and use B XJ/y(ee) data to determine their relative fraction»re-weight B + K + p + p - ee simulated events using background subtracted B + K + p + p - µµ data Candidates per 35 MeV/c 6 LHCb 5 4 3 1 (a) + B K + π + π - µ + µ - 1 15 m(k + π + π - 5 JHEP 1 (14) 64 ) [MeV/c ]
Fit Results ee Candidates per 34 MeV/c Candidates per 34 MeV/c Pulls 1 5 1 Signal 15 1 1 1 5 5 Combinatorial Signal * B X( YK )ee Combinatorial * B X( YK )ee 1 5 45 5 55 6 45 5 55 6 m(kπ m(kπee) [MeV/c ] 1 5 45 5 55 6 45 5 55 6 m(kπ m(kπee) [MeV/c ] 1 Candidates per 34 MeV/c Candidates Pulls per 34 MeV/c low-q central-q Candidates per 34 MeV/c Candidates Pulls per 34 MeV/c 1 8 7 3 1 6 5 1 4 1 3 1 1 5 1 1 35 3 1 5 1 15 1 5 1 1 5 45 5 55 6 45 5 55 6 m(kπ m(kπee) [MeV/c ] 5 Leakage due to brem tail of B K * J/y(ee) Signal Combinatorial Signal Λ b pkj/ ψ B s K Combinatorial * Λ b pkj/ψ J/ ψ * B s K J/ψ Signal Combinatorial Signal * B X( YK Combinatorial )ee * B K J/ ψ * B X( YK )ee B * K J/ψ
Yields Precision of the measurement driven by the statistics of the electron samples In total, about 9 and 11 B K * ee candidates at low- and central-q, respectively
Cross-Checks I Control of the absolute scale of the efficiencies via the ratio which is expected to be unity and measured to be Result observed to be reasonably flat as a function of the decay kinematics and event multiplicity Extremely stringent test, which does not benefit from the cancellation of the experimental systematics provided by the double ratio 3
Cross-Checks II BR(B K * µµ) in good agreement with [arxiv:166.4731] If corrections to simulations are not accounted for, the ratio of the efficiencies changes by less than 5% Further checks performed by measuring the following ratios which are found to be compatible with the expectations 4
Cross-Checks III Relative population of bremsstrahlung categories compared between data and simulation using B K * J/y(ee) and B K * g(ee) events Fraction of events [%] 6 5 4 3 1 * B K J/ψ(ee) (Data) * B K J/ψ(ee) (Simulation) Fraction of events [%] 8 7 6 5 4 3 1 * B K γ (Data) * B K γ (Simulation) not possible to assign unambiguously one photon to a track due to very small opening angle between electrons LE LH LI LE LH LI LE LH LI γ 1γ γ LE LH LI LE LH LI LE LH LI γ 1γ γ A good agreement is observed 5
Cross-Checks IV The splot technique is used to statistically subtract the background from the selected data [NIM A555, 356-369 (5)] Normalised distributions a.u..4.35.3.5..15.1.5 * B K µµ (Data) * B K ee (Data) * B K µµ (Simulation) * B K ee (Simulation).5 1 Normalised distributions a.u..4.35.3.5..15.1.5 low-q q [GeV /c 4 ] central-q * B K µµ (Data) * B K ee (Data) * B K µµ (Simulation) * B K ee (Simulation) 3 4 5 6 q [GeV /c 4 ] A good agreement is observed in both q regions between muons and electrons, data and simulation 6
Cross-Checks V No attempt is made to separate the K meson from S-wave or other broad contributions present in the mass peak region Normalised distributions a.u..45.4.35.3.5..15.1.5 * B K µµ (Data) * B K ee (Data) * B K µµ (Simulation) * B K ee (Simulation) 8 85 9 95 Normalised distributions a.u..45.4.35.3.5..15.1.5 low-q m(kπ) [MeV/c ] central-q * B K µµ (Data) * B K ee (Data) * B K µµ (Simulation) * B K ee (Simulation) 8 85 9 95 m(kπ) [MeV/c ] A clear K mass peak is visible, and the muon and electron channels manifest a very good agreement 7
Cross-Checks VI The opening angle between the two leptons Normalised distributions a.u..45.4.35.3.5..15.1.5..4.6.8.1 a.u..45.4.35.3.5..15.1.5 * B K µµ (Data) * B K ee (Data) * B K µµ (Simulation) * B K ee (Simulation).1..3.4.5 θ [mrad] ll.5.1.15. [mrad] The distribution is different between muons and electrons at low-q because of the difference in the lepton masses Even very close to threshold a good description is observed (insert,.45<q <.1 GeV /c 4 ) 8 Normalised distributions a.u..45.4.35.3.5..15.1.5 low-q [mrad] central-q θ ll * B K µµ (Data) * B K ee (Data) * B K µµ (Simulation) * B K ee (Simulation) θ ll
R K*º determined as a double ratio Systematics I» Many experimental systematic effects cancel» Statistically dominated (~15%) Total systematic uncertainty of 4-6% and 6-8% in the low- and central-q 9
Systematics II Corrections to simulation: besides the uncertainty due to the size of the samples, an additional systematic is determined using different parameterisations of the corrections Kinematic selection: a systematic uncertainty for Data/MC differences in the description of the bremsstrahlung tail and the MVA classifier is determined by comparing simulation and background subtracted B K * J/y(ll) data Residual background: both data and simulation are used to assess a systematic uncertainty for residual background contamination due to B K * J/y(ee) events with a K e or p e swap 3
Systematics III Mass fit: a systematic uncertainty is determined by running pseudo-experiments with different descriptions of the signal and background fit models Bin migration: the effect of the model dependence and description of the q resolution in simulation are assigned as a systematic uncertainty r J/y flatness: the ratio is studied as a function of several properties of the event and decay products, and the observed residual deviations from unity are used to assign a systematic uncertainty 31
Results I lnl 8 7 6 5 4 3 1 LL LH LI Combined Combined (stat).5 1 1.5 low-q * central-q R K.5 1 1.5 The measured values of R K*º are found to be in good agreement among the three trigger categories in both q regions lnl 8 7 6 5 4 3 1 LL LH LI Combined Combined (stat) 3 R K *
Results II 1.. R K 1..8 R K 1.5.6 1..4.. LHCb SM from CDHMV SM from EOS SM from flav.io SM from JC 1 3 4 5 6 q [GeV /c 4 ].5 LHCb BaBar Belle. 5 1 15 PRD 86 (1) 31 q [GeV /c 4 ] PRL 13 (9) 17181 The compatibility of the result in the low-q with respect to the SM prediction(s) is of.-.4 standard deviations The compatibility of the result in the central-q with respect to the SM prediction(s) is of.4-.5 standard deviations 33
Summary and Outlook Using the full Run 1 data set the R K*º ratio has been measured by LHCb with the best precision to date in two q bins The compatibility of the result with respect to the SM prediction(s) is of.-.5 standard deviations in each q bin The result is particularly interesting given a similar behaviour in R K Rare decays will largely benefit from the increase of energy (cross-section) and collected data (~5 fb -1 expected in LHCb) in Run LHCb has a wide programme of LU tests based on similar ratios Future measurements will be able to clarify whether the tantalising hints we are observing are a glimpse of NP 34
Backup
Calorimeter System Composed of a Scintillating Pad Detector (SPD), a Preshower (PS), an electromagnetic calorimeter (ECAL) and a hadronic calorimeter (HCAL) The SPD and the PS consist of a plane of scintillator tiles (.5 radiation lengths, but to only 6% hadronic interaction lengths) The ECAL has shashlik-type construction, i.e. a stack of alternating slices of lead absorber and scintillator (5 radiation lengths) The HCAL is a sampling device made from iron and scintillator tiles being orientated parallel to the beam axis (5.6 interaction lengths) 37
Cross-Checks III Relative population of bremsstrahlung categories compared between data and simulation using B K * J/y(ee) and B K * g(ee) events A good agreement is observed 38
Cross-Checks VII The distance between the Kp and ll vertices Normalised distributions a.u..6.5.4.3. * B K µµ (Data) * B K ee (Data) * B K µµ (Simulation) * B K ee (Simulation) Normalised distributions a.u..6.5.4.3. * B K µµ (Data) * B K ee (Data) * B K µµ (Simulation) * B K ee (Simulation).1.1 1 5 5 1 low-q z vtx -z vtx [mm] central-q ll Kπ 1 5 5 1 z vtx -z vtx ll [mm] Kπ The hadron and lepton pairs consistently originate from the same decay vertex 39
Results III What about NP? 1. R K 1..8.6.4.. LHCb CDHMV : C9µ NP = 1.1 CDHMV : C9µ NP = C1µ NP =.65 CDHMV : C9µ NP = C9 NP µ = 1.7 CDHMV : C NP 9µ = C NP 9 µ EOS: benchmark point C9µ NP = 1. EOS: data driven C9µ NP from P 5 and R K flav.io: C9µ NP = 1.1 flav.io: C9µ NP = C1µ NP =.65 = 1.18 and CNP 1µ = C NP 1 µ =.38 1 3 4 5 6 q [GeV /c 4 ] 4
Di-Lepton Mass 41
Theoretical Framework 4
Operators 43
Angular Analyses 44
Interpretation of Global Fits 45
Interpretation of Global Fits 46