BOOS24 University College London, 8-22 August 24 ALAS Measurements of Boosted Objects Christoph Anders on behalf of the ALAS collaboration Physikalisches Institut Universita t Heidelberg August 9th 24
Introduction and Content 2/2 Understanding behavior of basic boosted objects, W /Z and top, at high p is essential for many measurements and searches! New or updated ALAS cross section results since BOOS 3: Differential t t cross section (July 4) High p Z b b cross section (April 4) High p W /Z cross section (July 4)
Differential t t Cross Section Measurements 3/2 arxiv:47.37 [hep-ex], submied to PRD : s = 7 ev, L int = 4.6 fb Event selection: Single lepton trigger e or µ, p > 25 GeV, η < 2.5 4 anti-k t, R =.4 jets, p > 25 GeV, η < 2.5 b-tag, MVA tagger, ɛ = 7% E miss > 3 GeV and m W > 35 GeV Reconstruct t t kinematics with Likelihood fier Signal MC: ALPGEN+HERWIG, CEQ6L W +jets normalization and multi-jet backgrounds derived from data W +jets shape and other bkgs from MC
Differential t t Cross Section Measurements II 4/2 arxiv:47.37 [hep-ex], submied to PRD Events / GeV /Prediction 4 3 2-2 -3 ALAS s=7 ev L dt = 4.6 fb µ+jets t t t t (l+jets) (dilepton) Single top W+jets Multijet Other.5.5 2 3 4 5 6 7 8 Hadronic top p [GeV] Unfold data to parton-level kinematics Regularized SingularValueDecomposition method Migration matrices from t t MC (LO) Main systematic uncertainties: p t,m t t : JES (-4%), MC gen. (-8%), b-tagging eff. (-4%) p t t : ISR/FSR (%), MC gen. (-9%), frag. (-7%), JER (3-8%) y t t : MC gen. (-6%), fragmentation (-4%), Combine e and µ channels with BLUE method Compute σ dx, with X: p t (sensitive to higher order corrections @ high p ) m t t (important for t t resonance searches) p t t y t t (test of PDFs) (sensitive to extra radiation) Compare to various generators and calculations
t Comparison of Unfolded and Different Generators 5/2 σ dp t : GeV σ dm : t t σ dp MC GeV dp σ MC -3-4.5 ALAS L dt = 4.6 fb s = 7 ev ALPGEN+HERWIG MC@NLO+HERWIG POWHEG+HERWIG POWHEG+PYHIA.5 2 3 4 5 6 7 8 p t [GeV] σ dp t t : -2-3 -4-5 ALAS L dt = 4.6 fb s = 7 ev ALPGEN+HERWIG MC@NLO+HERWIG POWHEG+HERWIG POWHEG+PYHIA.5 2 3 4 5 6 7 8 9.5 2 3 4 5 6 7 8 9 p [GeV] GeV dm σ MC d y σ MC -3-4 -5 ALAS L dt = 4.6 fb s = 7 ev ALPGEN+HERWIG MC@NLO+HERWIG POWHEG+HERWIG POWHEG+PYHIA.2 5 5 2 25.8 5 5 2 25 m [GeV] σ dy : t t ALAS L dt = 4.6 fb s = 7 ev ALPGEN+HERWIG MC@NLO+HERWIG POWHEG+HERWIG POWHEG+PYHIA..5.5 2 2.5.9.5.5 2 2.5 y p t softer in data above 2 GeV. No single generator performs best for all kinematic variables considered. χ 2 tests show that the data have sufficient precision to probe the predictions!
Comparison of Unfolded and NLO QCD Calculations t 6/2 σ dp t : σ dm : t t GeV σ dp -3-4 ALAS NLO QCD L dt = 4.6 fb s = 7 ev GeV dm σ -3-4 -5 ALAS NLO QCD L dt = 4.6 fb s = 7 ev In data p t and m t t softer in tails NLO GeV dp σ NLO.5.5 2 3 4 5 6 7 8 p t [GeV] σ dp t t : -2-3 -4-5.5 ALAS NLO QCD L dt = 4.6 fb s = 7 ev.5 2 3 4 5 6 7 8 9 p [GeV] NLO d y σ NLO.5 5 5 2 25.5 5 5 2 25 m [GeV] σ dy : t t ALAS NLO QCD L dt = 4.6 fb.5.5.5 2 2.5 s = 7 ev.5.5.5 2 2.5 y Details: QCD calculations at NLO (MCFM with C PDF) µ F = µ R = m top, varied by 2 and 2 C PDF 68% CL error-sets
Comparison of Unfolded and NLO+NNLL QCD Calculations 7/2 t σ dp t : σ dm : t t GeV σ dp -3-4 ALAS NLO+NNLL L dt = 4.6 fb s = 7 ev GeV dm σ -3-4 ALAS NLO+NNLL L dt = 4.6 fb s = 7 ev In data p t and m t t softer in tails -5 GeV NLO+NNLL dp σ NLO+NNLL.5.5 2 3 4 5 6 7 8 p t [GeV] σ dp t t : -2-3 -4-5.5 ALAS NLO+NNLL L dt = 4.6 fb s = 7 ev.5 2 3 4 5 6 7 8 9 p [GeV] NLO+NNLL.5 5 5 2 25.5 5 5 2 25 m [GeV] Details: QCD calculations at NLO+NNLL (MSW28NNLO) Dynamic scale m t t, varied by 2 and 2 PDF uncertainties for m t t and p t t For p t dynamic scale uncertainty µ = mt 2 + Pt 2 included
Comparison of Unfolded and different PDFs 8/2 heory / heory /.8.7 ALAS.6.5.4.3.2..9 Ldt = 4.6 fb s = 7 ev σ dp t : CNLO MSW28NLO NNPDF 2.3 HERAPDF.5.8 5 5 2 25 35 8 t p [GeV] σ dp t t : 2.5 ALAS 2.5 Ldt = 4.6 fb s = 7 ev CNLO MSW28NLO NNPDF 2.3 HERAPDF.5.5 4 7 34 p [GeV] heory / heory /.8.7 ALAS.6.5.4.3.2..9 Ldt = 4.6 fb s = 7 ev σ dm : t t CNLO MSW28NLO NNPDF 2.3 HERAPDF.5.8 25 45 55 7 96 27 m [GeV] σ dy : t t.4 ALAS CNLO MSW28NLO.3 Ldt = 4.6 fb NNPDF 2.3 HERAPDF.5 s = 7 ev.2..9.8..5. 2.5 y MCFM predictions with 4 PDF sets Some tension at high p t and m t t might be usable to improve future PDF fits
High p Z b b cross section measurement 9/2 arxiv:44.742 [hep-ex], submied to Physics Leers B 3 n jet 5: anti-k t R =.4, p > 3 GeV, η < 2.5 n b jet = 2: MVA, ɛ = 7%, p > 4 GeV, R bb <.2 Di-jet system: p dijet = p bb > 2 GeV 6 < m djet < 6 GeV : s = 8 ev, L int = 9.5 fb Signal simulation: SHERPA Multi-jet background: from data Other small backgrounds from simulation. OR of six jet based triggers.
Separation from Multi-Jet Background Neural Network /2 η dijet : η(dijet, balance jet): ANN output S NN : ) dijet (/N)(dN/dη.35.3.25.2.5 ALAS s = 8 ev Ldt = 9.5 fb Z bb + (/N)(dN/d η).7.6.5.4.3 ALAS s = 8 ev Ldt = 9.5 fb Z bb..2.5..5.5 2 2.5 3 η dijet.5.5 2 2.5 3 3.5 4 4.5 5 η(dijet - BalanceJet) Z+jet system tends to be more boosted along beam axis than than multi-jet background. Use η dijet and η(dijet, balance jet) in NN. Use NN output S NN to split sample in signal and control region: 236k, Ndata CR 475k (S/B) SR = 6%, (S/B) CR = 2% for m dijet in [m W 5GeV, m W + 5GeV] NSR data Signal Region / Control Region.25.2.5..5.95.9.85.8 m dijet ratio SR/CR: ALAS s = 8 ev Ldt = 9.5 fb st Order Polynomial 6 7 8 9 2 3 4 5 6 m dijet [GeV] Very small correlation between NN and m dijet!
Signal Extraction /2 Events / 5 GeV 25 2 5 Signal Region: ALAS s = 8 ev Ldt = 9.5 fb Signal Region Signal + Background Background Fit: m dijet (6 GeV, 6 GeV) Simultaneous fit in SR and CR (binned ( GeV), extended maximum-likelihood). Events / 5 GeV 5 5 4 3 2 6 7 8 9 2 3 4 5 6 Control Region: ALAS s = 8 ev Ldt = 9.5 fb Control Region Signal + Background Background 6 7 8 9 2 3 4 5 6 Signal model: 3 Gaussians, yield in SR, peak position are free, other parameters fixed to MC. R Z = N CR Z b b /NSR =.62 fixed. Z b b Multi-jet background: 7 th order Bernstein polynomial Other smaller backgrounds from MC. Yield: N Z b b = 642 ± 64 (stat.)
Signal Extraction /2 ( - Background) / 5 GeV ( - Background) / 5 GeV 2 5 5-5 2 5 5-5 Signal Region: ALAS s = 8 ev Ldt = 9.5 fb Signal Region 6 7 8 9 2 3 4 5 6 Control Region: ALAS s = 8 ev Ldt = 9.5 fb Control Region 6 7 8 9 2 3 4 5 6 Fit: m dijet (6 GeV, 6 GeV) Simultaneous fit in SR and CR (binned ( GeV), extended maximum-likelihood). Signal model: 3 Gaussians, yield in SR, peak position are free, other parameters fixed to MC. R Z = N CR Z b b /NSR =.62 fixed. Z b b Multi-jet background: 7 th order Bernstein polynomial Other smaller backgrounds from MC. Yield: N Z b b = 642 ± 64 (stat.)
Resulting Cross Section σ Z b b 2/2 Measurement: σ Z b b(fiducial) = 2.2 ±.2 (stat.) ±.25 (syst.) ±.6 (lumi.) pb Systematics: NLO Predictions: σ POWHEG = 2.2 +.25.9 +.3 (scales).4 (PDF) pb σ amc@nlo =.98 +.6.8 (scales)±.3 (PDF) pb Very good agreement!
High p all-hadronic W /Z cross section measurement 3/2 arxiv:47.8 [hep-ex], submied to New J. Phys. : s = 7 ev, L int = 4.6 fb Signal MC: HERWIG+JIMMY scaled to NLO MCFM prediction (k =.25). QCD jet MC: PYHIA + variations for crosschecks/systematics. Reconstruct W /Z in one anti-k t R =.6 jet, with p > 32 GeV, η <.9, 5 GeV< m jet < 4 GeV Limited jet mass resolution measure σ W +Z = σ W (p > 32, η <.9) B (W q q) + σ Z (p > 32, η <.9) B (Z q q) Use W /Z enriched jet sample to study several grooming methods.
W /Z Jet- QCD Jet Discrimination 4/2 Boost jet to its center-of-mass frame: W /Z: back-to-back QCD: isotropic Analyse jet-shapes : Jet center-of-mass frame: hrust Minor: Sphericity: Aplanarity: Jets /.2.2..8 ALAS s = 7 ev, 4.6 fb p > 32 GeV η <.9 QCD jets W/Z jets 2 Jets /.4.3 ALAS QCD jets s = 7 ev, 4.6 fb W/Z jets.25 p > 32 GeV 2 η <.9.2 Jets /.2.5.4.3 ALAS s = 7 ev, 4.6 fb p > 32 GeV η <.9 QCD jets W/Z jets 2.6.4.5..2.2.5..2.4.6.8 hrust Minor.2.4.6.8 Sphericity..2.3.4.5 Aplanarity
Likelihood Discriminant 5/2 Maximise S/ S + B L >.5 Final sample: 59k jets n jet > for 2.5% of selected data events
Signal and Background Modeling 6/2 Signal: Background: Jets / 2 GeV 5 ALAS Simulation s = 7 ev p > 32 GeV η <.9 L >.5 W/Z jets otal signal W Z (a) Jets / 2 GeV 3 ALAS Simulation (b) 25 2 5 s = 7 ev p > 32 GeV η <.9 L >.5 QCD jets Background fit 5 5 6 8 2 4 Jet Mass [GeV] For each W and Z: Breit-Wigner Gauss Floating: W /Z rate Other parameters: fixed to MC 5 5 2 25 Jet Mass [GeV] 2 exp. decay functions + sigmoid S( m) = m/ + m 2 with m = ( m jet m ) /σm Shoulder due to selection, etc., well reproduced by QCD jet MC
Signal Extraction and Results 7/2 Jets / 2 GeV 25 2 5 5 2 Signal + Background fit Background fit component Signal fit component - Fit bkg 8 6 4 2 8 6 4 2-2 5 6 7 8 9 2 3 4 ALAS s = 7 ev, 4.6 fb Jet Mass [GeV] p > 32 GeV η <.9 L >.5 Systematics: 5 5 2 25 Jet Mass [GeV] Result: σ W +Z (W /Z q q, p > 32, η <.9) = 8.5 ±.8 (stat.) ±.5 (syst.) pb Consistent with NLO MCFM prediction within 2σ: σ MCFM W +Z = 5. ±.5 pb
Effects of Various Grooming echniques 8/2 Use selected sample enriched in W /Z to study effect of pruning (R cut =.3, z cut =.2, k t ) 2 trimming (R sub =.2, f cut =.3) 3 area subtraction & 2 : cut on recalculated LLH to get the same background rejection (89%). 3 : Recalculate area subtracted m jet for default jets. Events / GeV (a) ALAS 2 s = 7 ev, 4.6 fb Ungroomed jets p > 32 GeV, η <.9 Pruned jets rimmed jets Area Subtracted jets 8 6 4 2 PYHIA8(bkg)+HERWIG(sig) Ungroomed jets Pruned jets rimmed jets Area Subtracted jets 5 6 7 8 9 2 3 4 5 Jet Mass [GeV] Events / 2 GeV 9 8 7 6 5 4 3 2 (b) ALAS Simulation Ungroomed jets Pruned jets rimmed jets Area Subtracted Jets 5 6 7 8 9 2 3 4 5 Jet Mass [GeV] Without extra optimization, results are compatible with default analysis within statistical uncertainties.
Low vs. High Pileup, i.e. N Vtx < 5 vs. N Vtx > 9/2 Fraction of Jets / GeV.2..8.6 Ungroomed jets L >.5 Ungroomed: ALAS s p = 7 ev, 4.6 fb > 32 GeV η <.9 Fraction of Jets / GeV.2..8.6 Pruned Jets L pruned >.6 Pruned: ALAS s p = 7 ev, 4.6 fb > 32 GeV η <.9 Fraction of Jets / GeV.4 (a) : N Vtx > : N Vtx < 5.2 PYHIA8(bkg)+HERWIG(sig): N > Vtx PYHIA8(bkg)+HERWIG(sig):N < 5 Vtx 5 6 7 8 9 2 3 4 5.2..8.6 rimmed jets L trimmed >.6 rimmed: ALAS s p Jet Mass [GeV] = 7 ev, 4.6 fb > 32 GeV η <.9 Fraction of Jets / GeV.4 (b) : N Vtx > : N Vtx < 5.2 PYHIA8(bkg)+HERWIG(sig): N > Vtx PYHIA8(bkg)+HERWIG(sig): N < 5 Vtx 5 6 7 8 9 2 3 4 5.2..8.6 Area subtracted: Area Subtracted jets L >.5 ALAS s p Jet Mass [GeV] = 7 ev, 4.6 fb > 32 GeV η <.9.4 (c) : N Vtx > : N Vtx < 5.2 PYHIA8(bkg)+HERWIG(sig): N > Vtx PYHIA8(bkg)+HERWIG(sig): N < 5 Vtx 5 6 7 8 9 2 3 4 5 Jet Mass [GeV].4 (d) : N Vtx > : N Vtx < 5.2 PYHIA8(bkg)+HERWIG(sig): N > Vtx PYHIA8(bkg)+HERWIG(sig): N < 5 Vtx 5 6 7 8 9 2 3 4 5 Pile-up is nicely suppresed, effect are well described by MC. Jet Mass [GeV]
Conclusions 2/2 New differential t t cross section measurement @7 ev Comparison with MC generators, NLO(+NNLL) QCD computations and PDF sets softer at high p t than prediction, current reach: 9 GeV High p all-hadronic W /Z and Z b b cross section measurements in good agreeement with predictions est of substructure techniques in events enriched with hadronically decaying W /Z shows good agreement with prediction and nice pile-up suppresion. Looking forward to even more boosted Run-I/II measurements @ BOOS25!
Backup 2/2 Backup Backup Backup
t t Differential Cross Section χ 2 values 22/2
t t Differential Cross Section Systematic Uncertainties, Event Yields 23/2
t t t t Differential Cross Section Migration Matrices I 24/2 [GeV] Reconstructed p ALAS Simulation s=7 ev e+jets 8.%.%.%.%.6% 7.9% 84.7% 35.3%.4%.7% 2.% 6.9% 7.2% 6.2% 25.8%.% 2.5% 2.5% 54.4% 8.4% 2.% 2 3.% 4.6% 3.5% 53.5% 3.3% 4.5%.2% 5.6% 9.9% 53.9% 8.4% 7.5% 4.7% 2.5% 34.4% 58.8% 22.% 9.9% 5.3% 2.9% 2.2% 5 49.8% 5.% 7.% 3.5% 2.%.4%.% 5 5 2 25 35 8 t correlation:.85 Parton-level p [GeV] [GeV] Reconstructed p ALAS Simulation s=7 ev µ +jets 8.%.%.%.%.4% 7.9% 82.7% 35.2%.3%.6% 2.4% 6.% 7.5% 6.7% 25.7%.% 2.4% 2.4% 53.9% 8.2% 2.3% 2 2.8% 4.3% 3.5% 52.9% 4.5% 4.8%.7% 5.8% 9.6% 53.5% 8.3% 7.6% 3.6% 3.% 34.4% 59.% 22.9%.2% 5.5% 3.5% 2.6% 5 5.% 5.6% 7.% 3.6% 2.%.4%.% 5 5 2 25 35 8 t correlation:.84 Parton-level p [GeV] [GeV] Reconstructed m ALAS Simulation 27 95 7 55 45 25 25 s=7 ev e+jets.%.%.6% 7.% 75.2%.6%.6%.7% 67.2% 7.7% 3.5% 4.9% 59.9% 7.7% 4.4% 7.8% 55.7% 9.% 5.3%.8% 78.% 27.7% 8.7% 2.8%.% 45 correlation:.86 55 7 95 27 Parton-level m [GeV] [GeV] Reconstructed m ALAS Simulation 27 95 7 55 45 25 25 s=7 ev µ +jets.%.2%.6% 8.% 74.5%.6%.9% 2.% 66.% 7.5% 3.4% 4.6% 58.8% 7.4% 5.% 7.4% 55.4% 9.3% 5.6%.4% 78.5% 28.% 9.3% 2.8%.4% 45 correlation:.85 55 7 95 27 Parton-level m [GeV]
t t Differential Cross Section Migration Matrices II 25/2 [GeV] Reconstructed p ALAS Simulation 34 7 s=7 ev e+jets.%.% 3.% 72.7%.7% 4.3% 58.4% 2.% 3.7% 7.3% 35.% 6.% [GeV] Reconstructed p ALAS Simulation 34 7 s=7 ev µ +jets.%.% 3.% 72.3%.7% 4.% 57.4% 22.9% 3.6% 7.2% 35.9% 4.7% 4 4 68.6% 24.4% 3.6%.3% 68.7% 24.7% 3.6%.% correlation:.84 4 7 34 Parton-level p [GeV] correlation:.84 4 7 34 Parton-level p [GeV] Reconstructed y ALAS Simulation 2.5. s=7 ev e+jets.7%.4% 75.5% 4.% 68.4% 2.5% Reconstructed y ALAS Simulation 2.5. s=7 ev µ +jets.9%.6% 78.% 4.9% 68.8% 9.3%.5.5 85.3% 2.2% 3.% 84.3% 9.6% 2.7%.. correlation:.8.5. 2.5 Parton-level y.. correlation:.8.5. 2.5 Parton-level y
Z b b: Details 26/2
Fit Validation, p dijet raised from 2 to 25 GeV 27/2 Events / 5 GeV 8 7 6 5 ALAS s = 8 ev Ldt = 9.5 fb Signal Region dijet p > 25 GeV Signal + Background Background Events / 5 GeV 6 4 2 ALAS s = 8 ev Ldt = 9.5 fb Control Region dijet p > 25 GeV Signal + Background Background 4 8 3 6 2 4 2 6 7 8 9 2 3 4 5 6 6 7 8 9 2 3 4 5 6 ( - Background) / 5 GeV ALAS Signal Region dijet s = 8 ev p > 25 GeV 8 Ldt = 9.5 fb 6 4 2 ( - Background) / 5 GeV ALAS Control Region dijet s = 8 ev p > 25 GeV 8 Ldt = 9.5 fb 6 4 2-2 -4 6 7 8 9 2 3 4 5 6-2 -4 6 7 8 9 2 3 4 5 6 his yield measurement, extrapolated back to the baseline fiducial cross-section definition (i.e. with dijet p > 2 GeV), results in a consistent cross section of 2. +/-.28 (stat.) pb.
Fit Validation, only events selected by the dominant trigger 28/2 Events / 5 GeV 4 2 8 6 4 2 ALAS s = 8 ev Ldt = 9.5 fb Signal Region Dominant trigger Signal + Background Background Events / 5 GeV 3 ALAS Control Region s = 8 ev Dominant trigger 25 Ldt = 9.5 fb Signal + Background 2 5 5 Background 6 7 8 9 2 3 4 5 6 6 7 8 9 2 3 4 5 6 ( - Background) / 5 GeV 5 5 ALAS s = 8 ev Ldt = 9.5 fb Signal Region Dominant trigger ( - Background) / 5 GeV 5 5 ALAS s = 8 ev Ldt = 9.5 fb Control Region Dominant trigger -5-5 6 7 8 9 2 3 4 5 6 6 7 8 9 2 3 4 5 6 Statistically independent compared to next page, compatible results:.99 +/-.25 (stat.) pb and.87 +/-.44 (stat.) pb respectively, despite the very different background shapes that are produced by the different kinematic sculpting effects of the triggers and the conservative trigger modelling systematic uncertainty of the measurement.
Fit Validation, only events NO selected by the dominant trigger 29/2 Events / 5 GeV 9 ALAS Signal Region 8 s = 8 ev Other triggers Ldt = 9.5 fb 7 Signal + Background 6 Background 5 4 3 2 Events / 5 GeV 8 ALAS Control Region s = 8 ev Other triggers 6 Ldt = 9.5 fb 4 Signal + Background 2 8 6 4 2 Background 6 7 8 9 2 3 4 5 6 6 7 8 9 2 3 4 5 6 ( - Background) / 5 GeV 6 4 2-2 ALAS s = 8 ev Ldt = 9.5 fb Signal Region Other triggers ( - Background) / 5 GeV 6 4 2-2 ALAS s = 8 ev Ldt = 9.5 fb Control Region Other triggers -4 6 7 8 9 2 3 4 5 6-4 6 7 8 9 2 3 4 5 6 Statistically independent compared to last page, compatible results:.87 +/-.44 (stat.) pb and.99 +/-.25 (stat.) pb respectively, despite the very different background shapes that are produced by the different kinematic sculpting effects of the triggers and the conservative trigger modelling systematic uncertainty of the measurement.
W /Z: Event Shape Varaibles 3/2