Flavour agging at ALAS Hannah Arnold Albert-Ludwigs-Universität Freiburg Heavy Flavour Production at the LHC, IPPP Durham - April 6 H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 / 3
Introduction Introduction b-tagging = the identification of s containing b-hadrons important tool for background suppression in a wide range of analyses (e.g. top quark and Higgs sector, in the search for new phenomena) several dedicated algorithms with varying levels of complexity and performance exploiting the long lifetime, high mass and decay multiplicity of b-hadrons and the hard b-quark fragmentation reconstructable secondary vertices and distinct track properties performance characterized by power to separate b-, c- and light s in simulated events (requires flavour labelling based on simulation s truth record) challenges: dense and boosted environments, double b-hadron s,... for use in physics analyses measurements of b- tagging efficiency, c- tagging efficiency and mistag rate required data-to-mc scale factors (SF) eliminating / reducing dependence on simulations several (complementary) calibration methods developed combinations challenges: selecting samples with strong predominance of a single flavour, SF applicability to different simulations / processes,... H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 / 3
Introduction Outline Introduction b-agging algorithms 3 Jet truth flavour labelling 4 Calibration of b-tagging algorithms 5 Summary H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 3 / 3
b-agging algorithms b-agging algorithms b-agging algorithms lifetime-based impact parameter-based: (JetProb), IP3()D vertex-based: SV(), JetFitter combinations: IP3D+SV, IP3D+JetFitter, MV, MV (Run ) muon-based IP P V b hadron SV V b c hadron µ, e 3 b- trigger Key objects (small-r) calorimeter s axis used to define b-hadron flight path tracks reconstructed in the Inner Detector 3 signal primary vertex (PV) of HS collision (with tracks) tracks-to-calo. s association: based on angular separation R(track, ) R cut varies as a function of the p (decreasing with increasing p ) exclusive H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 4 / 3
b-agging algorithms Impact parameter(ip)-based algorithms I transverse IP d : distance of closest approach of track to PV in the r-φ projection longitudinal IP z : distance between z coordinates of the PV and the track at closest approach in r-φ bla improving separation power by: introducing sign: + ( - ) - track intersects the axis in front of (behind) the PV using significance, e.g. d /σ d tracks from b-/c-hadron decays: large d and z, + sign exp. resolution generates a random sign, tails at + from long-lived particles, conversions,... bla JetProb: relies on d significance; simple, robust, no simulation dependence Arbitrary units 3 4 5 ALAS Simulation Preliminary s=3 ev, tt b s c s Light flavour s 6 3 4 rack signed d significance (Good) H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 5 / 3
b-agging algorithms Impact parameter(ip)-based algorithms II transverse IP d : distance of closest approach of track to PV in the r-φ projection longitudinal IP z : distance between z coordinates of the PV and the track at closest approach in r-φ bla improving separation power by: introducing sign: + ( - ) - track intersects the axis in front of (behind) the PV using significance, e.g. d /σ d tracks from b-/c-hadron decays: large d and z, + sign exp. resolution generates a random sign, tails at + 3 from long-lived particles, conversions,... 4 bla IP3D: LLR-based, relies on d and z significances as 5 well as correlations 6 more powerful, D-PDFs from simulation IP3D log(p /P ) Arbitrary units ALAS Simulation Preliminary s=3 ev, tt b s c s Light flavour s 3 4 5 H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 6 / 3 b u
b-agging algorithms Vertex-based algorithms: SV algorithm II Explicit reconstruction of a single, inclusive secondary vertex (SV). using all associated, displaced tracks to form vertex candidates from track pairs (χ based) vertices compatible with long-lived particles and material interactions are rejected iterative procedure to combine all tracks from -track-vertices into single inclusive vertex small mistag rate SV finding efficiency: 7% bla SV: LLR-based, exploiting vertex mass, energy fraction, number of -track vertices, R(,PV-SV) Fraction of s /.8 3 4 p s=7 ev, tt > GeV, η <.5 ALAS Simulation b s c s light-flavour s 5..4.6.8 SV energy fraction H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 7 / 3
b-agging algorithms Vertex-based algorithms: JetFitter algorithm Explicit reconstruction of the complete b-hadron decay chain. exploiting topological structure of weak b- and c-hadron decays inside uses Kalman filter to find common line between and position of PV, SV and V (tertiary, c-hadron decay vertex) approximating b-hadron flight path one track sufficient to built vertex! six variables: decay topology + vertex information bla input for artificial neural network Fraction of s /.4 ALAS Simulation, b s. single track single tracks. tracks.8 3 or more tracks three output nodes corresponding to the b-,c- and light hypotheses P b/c/l.6.4. s=7 ev, tt p > GeV, η <.5...3.4.5.6.7.8.9 final discriminating variables: ln(p b/p l) (JetFitter), ln(p b/p c) (JetFitter(c)) Energy fraction H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 8 / 3
b-agging algorithms Performance I figure of merit: b- tagging efficiency vs. light (c-) rejection defined by placing cuts on b-/c/-light distributions of the disciminating variables in simulated t t events clear hierarchy between standalone and combined algorithms Fraction of s /. 3 4 ALAS Simulation s=7 ev, tt b s c s light-flavour s p > GeV, η <.5 5 MV = IP3D+SV+(IP3D+JetFitter) (NN)..4.6.8 MV weight Light-flavour- rejection 4 3 MV IP3D+JetFitter IP3D+SV JetFitter IP3D SV JetProb c- rejection IP3D+JetFitter(c) IP3D+JetFitter MV ALAS Simulation s=7 ev, tt p > GeV, η <.5 ALAS Simulation s=7 ev p > GeV, η <.5.3.4.5.6.7.8.9 b- tagging efficiency.3.4.5.6.7.8.9 b- tagging efficiency H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 9 / 3
b-agging algorithms Performance II: Run improvements Improved performance in Run due to upgrade of the Inner Detector: added fourth pixel layer, Insertable B-Layer (IBL) improved track reconstruction (shared hits at high p ) revisited b-tagging algorithms: MVcxx BD combining 4 input variables from SV, IP()3D, JetFitter algorithms xx = c- fraction in simulated t t events used for training Light flavour rejection 4 3 ALAS Simulation Preliminary MVc MVc c rejection ALAS Simulation Preliminary MVc MVc p s=3 ev, tt > GeV, η <.5.5.55.6.65.7.75.8.85.9.95 b efficiency p s=3 ev, tt > GeV, η <.5.5.55.6.65.7.75.8.85.9.95 b efficiency small admixture of c-s only slightly degrades light- reject, significantly improves c- rejection H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 / 3
b-agging algorithms Performance III: Run improvements Improved performance in Run due to upgrade of the Inner Detector: added fourth pixel layer, Insertable B-Layer (IBL) improved track reconstruction (shared hits at high p ) revisited b-tagging algorithms: MVcxx BD combining 4 input variables from SV, IP()3D, JetFitter algorithms xx = c- fraction in simulated t t events used for training Light flavour rejection Light flavour rejection 5 4 4 3 3 Run / Run 6 5 4 s=8,3 ev, t p >5 GeV, η <.5 t ALAS Simulation Preliminary MVc Run MVc Run 3.55.6.65.7.75.8.85.9.95.55.6.65.7.75.8.85.9.95 b b efficiency c rejection c rejection Run / Run.5.5 s=8,3 ev, t p >5 GeV, η <.5 t ALAS Simulation Preliminary MVc Run MVc Run.5.55.6.65.7.75.8.85.9.95.55.6.65.7.75.8.85.9.95 b b efficiency @7% b-tagging eff.: factor 4 (.5-) improved light-(c-) rejection! H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 / 3
b-agging algorithms Performance IV: differential Efficiency ALAS Simulation Preliminary s=3 ev, tt Efficiency ALAS Simulation Preliminary s=3 ev, tt Efficiency ALAS Simulation Preliminary s=3 ev, tt MVc ε =7% b b s c s Light flavour s MVc ε =7% b b s c s Light flavour s MVc ε =7% b b s c s Light flavour s 3 3 4 5 6 Jet p [GeV] 3.5.5.5 Jet η 3 5 5 5 3 35 4 Average interactions per bunch crossing b-tagging depends on p and η, as well as average number of pile-up interactions, different impact on b-/c- and light- efficiencies degradation of b- tagging efficiency at very low p : the b-hadron flight path is (too) short, tracks are (too) soft and not very displaced high p : merging of tracks, detector inefficient for very late decays, increased track multiplicity due to fragmentation H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 / 3
b-agging algorithms Performance V: dense environment and boosted topologies Scenario: searches for new physics at high masses highly boosted and collimated decay products, e.g. h b b R=.4 calo. s might not be able to resolve the products from the two b-hadron decays large-r (e.g..) calo. s: Higgs apply b-tagging to ghost-associated (R=.) track s small R parameters possible low PU sensitivity good angular resolution also in dense environment exploit large-r substructur to improve bkg. rejection? Arbitrary Units.5.45.4.35.3.5..5..5 ALAS Simulation Preliminary anti-k t R=. s rimmed (f =.5, R =.) cut sub 35 < p < 5 GeV, η <. det Double b-tagged @ 7% WP Higgs- Multi- Hadronic top 5 5 5 Multi rejection Normalized to unity. ALAS Simulation Preliminary s = 8 ev, M =. ev, k/m rack s (R=.) P =. anti-k rack s (R=.3) t track s k t calo subs Calo subs (R=.3) rack s (R=.4).8.6.4. 3 4 Num b-flav labeled s...3.4.5.6.7.8.9 Mass [GeV] Higgs efficiency H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 3 / 3 7 6 5 4 3 ALAS Simulation Preliminary anti k t R=. s rimmed (f =.5, R =.) cut sub 35 < p < 5 GeV, η <. det 68% mass window Double b tagging β= D + ( b tags at 7%) β= C + ( b tags at 7%) τ + ( b tags at 7%) wta τ + ( b tags at 7%)
b-agging algorithms Performance VI: double b-hadron s b-tagging algorithms currently optimized on t t samples, i.e. for s containing a single b hadron no information on the number of b-hadrons competing effects if two b-hadron decays inside one ( SVs, higher multiplicity of displaced tracks,...) likely improve (decrease) performance of IP-based and SV taggers (JetFitter) possible strategies to separate single and double b-hadron s: excplicitely reconstruct two b-hadron decay vertices / chains exploit substructure, kinematic differences Arbitrary Units...8.6.4...8.6.4. < GeV 8GeV < Jet P ALAS Preliminary Simulation Jet η <. single b-s merged b-s 5 5 5 3 Jet track multiplicity Merged b- rejection 4GeV<Jet p <6GeV 6GeV<Jet p <8GeV 8GeV<Jet p <GeV GeV<Jet p <5GeV 5GeV<Jet p <GeV GeV<Jet p <7GeV 7GeV<Jet p <36GeV 36GeV<Jet p <48GeV ALAS Preliminary Simulation.3.4.5.6.7.8.9 b- efficiency Enriched b- sample using MV@6%. Other MVA inputs: track- width, distances between tracks / subs H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 4 / 3
Jet truth flavour labelling Jet truth flavour labelling Run select partons with p > 5 GeV cone-based matching to reco. : R <.3 Run select weakly decaying hadrons with p > 5 GeV exclusive cone-based matching to reco. : R <.3 alternative: ghost-association Order of labelling if b parton(hadron) is found b else if c parton(hadron) is found c else if tau is found tau else light Comments several analyses already used Run baseline in Run cone-based matching and ghost-association agree at the % level in case of b-s (in t t) difference has non-negligible impact on light- rejection cone-based labelling coherent with track- association baseline! g b b/c c s are labelled as b-/c-s agrees with b-tagging point of view H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 5 / 3
Calibration of b-tagging algorithms Calibration of b-tagging algorithms the performance of b-tagging algorithms is optimized and evaluated using simulations, usually of inclusive t t events cannot expect simulations to describe all effects (modelling, detector) that impact the performance of b-tagging algorithms accurately measurements of the b-/c- and light- tagging efficiencies needed; as function of p and η requires extracting samples of s dominated by a single flavour results are presented in terms of data-to-simulation scale factors SF = ε data x /ε MC x, x = b, c, light assumption: SFs are process independent i.e. SF measured in a t t dilepton sample, applicable to a W+s sample SFs are MC generator dependent current approach: MC-to-MC SFs better: derive SFs for different MC generators H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 6 / 3
Calibration of b-tagging algorithms Calibration methods: an overview b- tagging efficiency calibration muon-based methods (di samples) p rel system8 t t-based methods (single-/dilepton t t) inclusive! tag counting kinematic selection kinematic fit combinatorial likelihood most precise! combined with muon-based methods c- tagging efficiency calibration W+c method (soft-muon tagged) D method (D D ( Kπ)π) mistag rate calibration: negative tag method bla Differences: purity, inclusiveness, simulation dependence, precision Jets /.5 GeV 45 ALAS dilepton 4 - L dt = 4.7 fb OS-SS Events /. 35 3 5 5 5 5 s = 7 ev data tt single-top diboson Z + s fake leptons norm. unc. 5 5 5 3 35 4 45 Jet p [GeV] 35 ALAS s = 7 ev Data W+c 3 - W+light 5 L dt = 4.6 fb Multi Others 5...3.4.5.6.7.8.9 MV weight H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 7 / 3
Calibration of b-tagging algorithms Calibration results c- tagging efficiency scale factor.5.5.5 b- tagging efficiency scale factor..8.6 MV, ε b = 7% D* ALAS - L dt = 4.7 fb s = 7 ev Combined (stat+syst) Combined (stat) Combinatorial likelihood on tt Di 3 4 5 6 7 8 9 Scale factor (stat) Scale factor (stat+syst) Inclusive c- sample ALAS s = 7 ev - L dt = 4.7 fb 4 6 8 4 Jet p [GeV] Mistag rate scale factor.5.5.5 = 7% MV, ε b 3 4 5 3 Jet p [GeV] - ALAS L dt = 4.7 fb, s = 7 ev negative tag Scale factor (stat) Scale factor (stat+syst) = 7% MV, ε b η <. otal uncertainties for MV@7%: b- calib. -8% (comb. like.), c- calib. 8-5% (D ), otal uncertainties for MV@7%: light- calib. 5-4 % H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 8 / 3 Jet p [GeV]
Calibration of b-tagging algorithms Calibration results: dense invironments I b tagging efficiency..9.8.7.6.5 ALAS Preliminary L dt=.3 fb s= 8 ev tt SL &P (stat.) stat. + syst. unc. Simulation MV@7%.4.6.8..4.6.8 min Jet R top: b- tagging efficiency for s from the hadronic top decay in single-lepton t t events as function of left: the minimal distance to a close-by right: the distance between the axis and the PV-SV axis SFs: and flat across p b tagging efficiency...9.8.7 ALAS Preliminary L dt=.3 fb s= 8 ev tt SL &P (stat.) stat. + syst. unc. Simulation MV@7%.5..5..5 R(vertex,) bottom: comparison of SFs with s from the leptonic top decay compatible Jet p [GeV] /ε sim. b data ε b.4.3...9 ALAS Preliminary L dt=.3 fb, s= 8 ev tt hadronic &P (stat.) tt leptonic &P (stat. + syst. unc.) stat. + syst. unc. PDF (stat. + syst. unc.) 5 5 5 3 35 4 45 5 tt MV@7% H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 9 / 3
Calibration of b-tagging algorithms Calibration results: dense invironments II Double b-tagging Rate Data / MC.8.6.4...8.6.4. ALAS Preliminary - s=8 ev,.3 fb anti-k t R=. rimmed (R =.3, f cut =.5) sub Data Pythia 8 MC.9 otal Uncertainty Fit Uncertainty b-tagging Uncertainty.6 p [GeV].3.7.4 3 4 5 6 7 8 9 [GeV] p Efficiency for two b-tagged (MV@7%) track s ghost-associated to large-r calo.. H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 / 3
Calibration of b-tagging algorithms Impact of (heavy flavour) modelling on b-tagging efficiencies calibration s goal: resolve (minimize) simulation dependence calibration s goal: (including detector response, reconstruction,...) calibration methods exploiting inclusive samples are doing a good job still rely on simulation for e.g. (ratios of) flavour fractions, distributions of discriminant variables,... accounted by applying dedicated uncertainties modelling uncertainties dominant for certain methods and phase space regions can be resolved by refined, more data-driven methods, e.g. combinatorial likelihood method more difficult in case of methods relying on non-inclusive samples: SFs applicable to inclusive samples if the differences of properties (affecting b-tagging) between the non-inclusive and the inclusive sample are well described ε = α ε excl SF = αdata α MC SF excl if α data = α MC SF = SF excl α data = α MC? If not need to extrapolate, i.e. estimate α data α MC. c- tagging efficiency ε c ε /ε c c(µ).9.8.7.6.5.4.3...8 ALAS Simulation Preliminary Pythia-default Incl. c- sample SM c- sample 85% 75% 7% 6% b-tagging operating point ε b H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 / 3
Calibration of b-tagging algorithms Extrapolation to inclusive samples goal: estimate α data from simulation where heavy quark fragmentation and heavy hadron decay properties are corrected to best knowledge: α data α corr MC SF αcorr MC α MC SF excl ratio SF extrapolation relatively independent from reference simulation Example: corrections applied in W+c calibration (reference MC: Pythia 6) the fragmentation fractions (e + e and e ± p data) the total branching ratio of the semileptonic decay of c-hadrons (PDG) the branching ratios of some exclusive semileptonic decays (PDG) the topological branching ratios of hadronic n-prong decays (PDG / EVGEN) the momentum of the decay muon in the rest frame of the c-hadron p (EVGEN) (the impact of the c-quark fragmentation function was evaluated, but not corrected) H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 / 3
Calibration of b-tagging algorithms Extrapolation to inclusive samples goal: estimate α data from simulation where heavy quark fragmentation and heavy hadron decay properties are corrected to best knowledge: α data α corr MC SF αcorr MC α MC SF excl ratio SF extrapolation relatively independent from reference simulation Example: corrections applied in W+c calibration (reference MC: Pythia 6) the production fractions (e + e and e ± p data) the total branching ratio of the semileptonic decay of c-hadrons (PDG) today: EVGEN*! the branching ratios of some exclusive semileptonic decays (PDG) today: EVGEN*! the topological branching ratios of hadronic n-prong decays (PDG / EVGEN) does not the topological branching ratios of hadronic n-prong decays (PDG / EVGEN) cancel! the momentum of the decay muon in the rest frame of the c-hadron (EVGEN) (the impact of the c-quark fragmentation function was evaluated, but not corrected) *with ALAS decay table H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 3 / 3
Calibration of b-tagging algorithms Heavy hadron production fractions Production fraction.5.4.3 ALAS Simulation Preliminary Pythia8 PYHIA6 Herwig++ Production fraction.7.6.5.4 ALAS Simulation Preliminary Pythia8 PYHIA6 Herwig++. HERWIG.3 HERWIG... B + B B s + B c Λ b - - Ξ b Ξ Ω b b Σ b Σ b Weakly decaying B hadrons - + Σ b + D D + D s + + Λ c Ξ c Ξ c Ω c Weakly decaying C hadrons large differences in production fractions between generators existing measurements for b-hadrons: HFAG (evatron), HFAG (evatron + LEP + LHCb) for c-hadrons: several measurements in e + e and e ± p data combination differ among each other and with simulation (Pythia 6 is rather close!) impact on b-tagging efficiencies is rather small; more pronounced for c-hadrons whose tagging efficiencies are rather different for MV@7%: ε(d + )/ε(d ).8 for MV@7%: ε(b + )/ε(b ).3 H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 4 / 3
Calibration of b-tagging algorithms Heavy hadron branching fractions inclusive and prominent exclusive semi-leptonic decays of dominant, weakly decaying heavy hadrons measured to high precision implemented in EVGEN via ALAS dec. table! (other generators show large differences) many exlcusive hadronic decays are measured, but still make up only a relatively small fraction of all hadronic decays e.g. D + : 6%, b-baryons: much less large differences between the number of implemented exclusive hadronic decays in different generators (EvtGen» Pythia8» Pythia6» Herwig) and their branching fractions BU: mainly the number of charged decay products has an impact on the tagging efficiency - unfortunately here the situation is not much better H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 5 / 3
Calibration of b-tagging algorithms Inclusive charged particle multiplicities only for the D the topological (inclusive) n-prong fractions are measured from which one can infer the hadronic ones... the mean of inclusive charged particle multiplicities is measured for the admixture of B, B +, B s, Λ b: EvtGen (4.76) «4.97 ±.7 «Pythia 6 (5.) BU: b-tagging efficiency depends on spectrum - large differences between generators Fraction of decays.5.4.3 ALAS Simulation Preliminary Pythia8 PYHIA6 Herwig++ HERWIG Pythia8+EvtGen Fraction of decays.8.7.6.5.4 ALAS Simulation Preliminary Pythia8 PYHIA6 Herwig++ HERWIG Pythia8+EvtGen..3... 5 5 5 3 + Multiplicity of B charged stable decay products 4 6 8 4 6 + Multiplicity of D charged stable decay products dominant (extrapolation) uncertainty in p rel, W+c and D calibrations! H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 6 / 3
Calibration of b-tagging algorithms Using EVGEN... b tagging efficiency.85.8.75.7 ALAS Simulation Preliminary tt simulation, s=8 ev p > GeV, η <.5 MV b tagging efficiency.85.8.75.7 ALAS Simulation Preliminary tt simulation, s=8 ev p > GeV, η <.5 MV.65.6.55 Pythia 6 Pythia 8 Herwig Herwig++.65.6.55 Pythia 6 (+EvtGen) Pythia 8 (+EvtGen) Herwig (+EvtGen) Herwig++ (+EvtGen) others / Pythia 6.5.95 5 5 5 3 p [GeV] others / Pythia 6.5.95 5 5 5 3 p [GeV] harmonizes heavy hadron decays reduces the differences in the b- tagging efficiency predicted by generators significantly; only small differences at the % level, mainly at low p remain harmonization less pronounced in case of the c- tagging efficiency - fragmentation fractions, tracks from fragmentation, etc. have a stronger impact H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 7 / 3
Calibration of b-tagging algorithms Summary on HF modelling using EVGEN harmonizes the b- tagging efficiencies significantly, but cannot resolve the simulation dependency of SFs completely (especially for c-s), since only affects heavy flavour decay modelling other corrections are still needed when extrapolating SFs to inclusive samples only comparisons between different generators allow to assess modelling uncertainties other than semi-leptonic branching fractions theoretically varying those within their measured uncertainties would suffice, but more complicated to do the estimated SF extrapolation factors (rel. unc.) are for MV: p rel : (3%); W+c:.86.95 (3-7%); D :.8.9 (the corrected eff. extrapolation factors are W+c:.7.75 «D :.5.6) H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 8 / 3
Summary Summary several dedicated algorithms to identify s containing b-hadrons high performing combined algorithm further improved for Run several methods to calibrate b-, c- tagging efficiencies and light- mistag rate using EVGEN consistently to decay heavy hadrons in Run simulations reduced generator dependency of the b- tagging efficiency signficantly b-tagging of small-r track s ghost-associated to large-r calo. s improves b-tagging in dense and boosted topologies work ongoing to better exploit substructure and to improve single and double b-hadron separation H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 9 / 3
Backup BACKUP H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 3 / 3
Backup References link title 5.94.pdf Performance of b-jet Identification in the ALAS Experiment ALAS-CONF--.pdf Identification and agging of Double b-hadron s with the ALAS Detector ALAS-CONF-3-9.pdf Calibration of the b-tagging efficiency for c s with the ALAS detector using events with a W boson produced in association with a single c quark ALAS-CONF-6-.pdf Calibration of ALAS b-tagging algorithms in dense environments ALAS-CONF-6-.pdf Studies of b-tagging performance and substructure in a high p g b b rich sample of large-r s from pp collisions at s = 8 ev with the ALAS detector AL-PHYS-PUB-4-3.pdf Flavor agging with rack Jets in Boosted opologies with the ALAS Detector AL-PHYS-PUB-4-4.pdf b-tagging in dense environments AL-PHYS-PUB-5-.pdf Expected performance of the ALAS b-tagging algorithms in Run- AL-PHYS-PUB-5-35.pdf Expected Performance of Boosted Higgs ( b b) Boson Identification with the ALAS Detector at s = 3 ev AL-PHYS-PUB-6-4.pdf Simulation of top quark production for the ALAS experiment at (s) = 3 ev AL-PHYS-PUB-4-8 Comparison of Monte Carlo generator predictions for bottom and charm hadrons in the decays of top quarks and the fragmentation of high p s H. Arnold (Universität Freiburg) Flavour agging at ALAS - April 6 3 / 3