tt+jets simulation comparisons Kenneth Wraight tt+jets simulation comparisons Alpgen&MC@NLO generator level MC@NLO Fast&Full simulation
semileptonic ttbar decays Looking for 4 jets of interest: 2 b quarks and 2 hadronic W daughters Trigger on lepton: e (>25GeV), μ (>20GeV). (↑Tevatron’s favourite (~85%)) LHC: ~95% gluon-gluon production.
Study top quark properties: charge, spin, etc. Motivations Study top quark properties: charge, spin, etc. Background to associated Higgs production and multi-jet SUSY decays (~1 ttbar event per second for σ=833pb at L=1033cm-2s-1): must be controlled. Ultimate test of generation and simulation software: multiscale QCD calculations
exact predictions for fixed orders First Comparison: MC@NLO & Alpgen Generator level MC@NLO: exact predictions for fixed orders higher order MEs with NLO accuracy to describe inclusive rates and LO accuracy for +1 jet Alpgen: approximate predictions for all orders (LL accuracy for FS + N jets) Consistent merging of LO MEs with shower MCs for multiparton FSs MC@NLO ~25k 5200 sample (red): Fully hadronic removed (es&μs) Default cuts Alpgen ~50k homemade (blue): Semileptonic (es only) ptjet ≥ 20GeV ηjet ≤ 6.0, ηbjet ≤ 5.0, ηcjet ≤ 5.0 ΔRjet ≥ 0.7 plots area normalised...
Pt & η of generator level tops & Ws
Gen level bs & W daughters kinematics
Comparing Full Simulation and Fast Simulation Second Comparison: MC@NLO Fast & Full Simulation Comparing Full Simulation and Fast Simulation Atlfast ~4-5 orders of magnitude faster than Full Simulation Replaces detector simulation and jet reconstruction part of simulation chain with smears on MC truth info Matching GL to Simulation: Begin by finding partons of interest in GL using mother & daughter information Match to shower jets using ΔR≤0.4 (take lowest value) veto any double matches repeat procedure for shower to simulation jets. MC@NLO ~25k 5200 sample Fullsim Pjets (red) Fastsim AFjets (blue) …no longer area normalised.
Relative finishing positions of matched b and W daughter jets Pt-Ordered position within the event of the highest Pt b-quark Pt-Ordered position within the event of the highest Pt W daughter
Pt of matched b and W daughter jets Pt spectrum of b-quark with highest Pt Pt spectrum of W daughter with highest Pt
Other jet relative finishing positions and pt 91 corresponds to cluster fragments
Reconstruction recipe poi 1st 2nd 3rd 4th 5th 6th other W1 28.79 28.90 29.07 29.04 24.44 24.38 12.55 12.50 3.81 3.89 1.10 1.12 0.24 0.18 W2 1.04 1.04 5.61 5.62 16.08 16.05 30.77 31.47 23.14 24.52 10.75 12.68 12.62 8.63 b1 48.56 48.39 28.91 28.88 13.49 13.47 6.36 6.50 2.11 2.19 0.36 0.42 0.22 0.14 b2 4.89 4.86 22.25 22.12 25.86 25.83 21.95 22.24 12.75 13.61 5.71 7.01 6.58 4.31 ¬poi 38.11 32.80 6.38 5.52 5.40 4.71 7.68 6.65 9.66 8.95 12.34 12.19 20.44 29.19 For triggered events, use jets within η≤3.2, ptjet>20GeV Cut on 5 hardest jets to reduce combinatoric background Reconstruct W using di-jet combinations Uses generator level info for b-vetos Reconstruct top using di-jets in W mass peak to make tri-jet combinations by adding btagged jets Simple reconstruction i.e. no calibration.
Fitted top & W reconstructed masses (with ideal btagging) Fullsim: W: mass 78036.8, signal 946.53, background 588.601 top: mass 164647, signal 928.002, background 1050.76 FastSim: W: mass 73479.4, signal 1078.14, background 459.972 top: mass 159449, signal 906.317, background 743.803 Fitted top & W reconstructed masses (with ideal btagging)
Fitted P top & W masses and pt spectra with realistic btagging Using fullsim b-weights cut=4.0 Fitted P top & W masses and pt spectra with realistic btagging W: mass 78390.1, signal 704.917, background 1147.25 top: mass 163346, signal 787.559, background 993.245 with additional pt<150GeV cut on top candidates: top: mass 168036, signal 650.501, background 538.357
Extra jet Extra jet pt: Jets from “well-reconstructed top” events: b-vetoed and not part of top tri-jet
Alpgen partons are slightly harder at generator level Summary Presented: Alpgen partons are slightly harder at generator level Fullsim & Fastsim comparable for MC@NLO Fullsim provides higher mass W&top on reconstruction So Far: Compared Alpgen & MC@NLO semileptonic tt+jets samples All justify cut on 5 hardest jets top reduce combinatorics similar study possible for semileptonic associated Higgs production To do: Crosscheck and compare with Sherpa Look into calibration and reconstruction efficiencies.
ALPGEN: Finishing positions Back up Alpgen slides... FastSim MC@NLO (blue), FastSim Alpgen (pink) poi 1st 2nd 3rd 4th 5th other W1 30.73 28.90 28.70 29.04 25.24 24.38 10.1412.50 3.39 3.89 1.8 1.30 W2 1.07 1.04 6.31 5.62 18.05 16.05 35.57 31.47 19.70 24.52 19.30 21.31 b1 48.38 48.39 30.36 28.88 13.04 13.47 5.34 6.50 1.85 2.19 1.04 0.56 b2 5.05 4.86 22.39 22.12 27.66 25.83 23.35 22.24 11.69 13.61 9.86 11.32 ¬poi 33.35 32.80 4.93 5.52 4.08 4.71 5.34 6.65 8.54 8.95 43.77 41.38 Cut on 5 hardest jets is justified!
ALPGEN: Gen Level W daughters matched to AF jets Pt-ordered position within the event of the highest Pt W daughter Pt spectrum of W daughter with highest Pt Pt spectrum of Other W daughter
ALPGEN: Gen Level b partons matched to AF jets Pt-ordered position within the event of the highest Pt b-quark Pt spectrum of b-quark with highest Pt
ALPGEN: Gen Level partons matched to AF jets Other jets’ finishing positions: First-finishing other jets seem to originate form quarks(1-5), gluons(21), photons(22) and cluster fragmentation(91).
Extra jet Extra jet pt: Jets from “well-reconstructed top” events: b-vetoed and not part of top tri-jet Of the original 53,286 event inclusive sample (comparable to ~43,000 ttbar event in first 2 months): 18603 matched from Gen.Level to AtlFast (~35%) 13867 were reconstructed (~26%)