1 / 32

Update on the hadronic top squark search: Top reconstruction @ 100 TeV

Update on the hadronic top squark search: Top reconstruction @ 100 TeV. Loukas Gouskos , Allan Sung, Joe Incandela. Physics motivation. SUSY: one of the most extensively studied BSM theories An excellent answer to: hierarchy problem, Dark Matter, unification of couplings

Download Presentation

Update on the hadronic top squark search: Top reconstruction @ 100 TeV

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Update on the hadronic top squark search: Top reconstruction @ 100 TeV Loukas Gouskos, Allan Sung, Joe Incandela FCC-hh analysis meeting

  2. Physics motivation • SUSY: one of the most extensively studied BSM theories • An excellent answer to: hierarchy problem, Dark Matter, unification of couplings • Top squark searches @ 100 TeV: • Understand naturalness of the EWK state • Viability of SUSY [top squarks should be < ~10 TeV] Current status Papucci et al. hep-ph 1110.6926 Recent articles relax these conditions: Mstop<~3 TeV[e.g. Baer et. al. 1602.0769, 1611.08511Ross et. al. 1110.6926] ̴TeV FCC-hh analysis meeting

  3. Physics motivation • SUSY: one of the most extensively studied BSM theories • An excellent answer to: hierarchy problem, Dark Matter, unification of couplings • Top squark searches @ 100 TeV: • Understand naturalness of the EWK state • Viability of SUSY [top squarks should be < ~10 TeV] Current status Papucci et al. hep-ph 1110.6926 Recent articles relax these conditions: Mstop<~3 TeV[e.g. Baer et. al. 1602.0769, 1611.08511Ross et. al. 1110.6926] GOAL: Probe the ~10 TeV regime with FCC-hh ̴TeV FCC-hh analysis meeting

  4. Signal characteristics • Top squark decays: very distinct signature in the all hadronic channel • Key player of top squark searches: top tagging • Top quarks are typically boosted decay products merge into a single jet • Identification of boosted top quarks provides a powerful handle to suppress many of the SM backgrounds • Multiple jets • 2 b-jets • On-shell top quarks • Large MET [from the two LSPs] Mj~Mt b W q1 t q2 FCC-hh analysis meeting

  5. Object reconstruction & event selection FCC-hh analysis meeting

  6. Objects & Baseline selection • Baseline selection: • Veto events with electrons & muons • Nj (high-pT)>=2 & Nb >= 1 • MET> 2 TeV • Δφ (j1,2 ; MET) > 0.5 ; Δφ (j3 ; MET) > 0.3 • low-pT jets: • PF Jets, R=0.4 • pT> 20 GeV • high-pT jets: • PF Jets, R=0.4 • pT> 1000 GeV • b jets: • PF Jets, R=0.4 • pT> 20 GeV • Electrons / Muons: • pT> 30 GeV • Iso/pT < 0.4 FCC-hh analysis meeting

  7. Background composition • Relevant backgrounds after the baseline sections: • QCD multijet • Huge cross-section • Rejected with large met and top-tagging selection • EWK backgrounds: Z/W+jets, VV, VVV • Leptonic decays where leptons escape lepton vetoes -> large MET • Highly suppressed after applying top-tagging and requiring multiple b-jets • Top (+X) backgrounds: ttV, ttVV, tttt, single-t • similar characteristics as signal • Multiple tops, multiple bs and MET from invisible particles • main challenge: tt & ttZ(->nunu) : needs dedicated analysis techniques [next slides] Prod xsec @ 100 TeV [M Mangano et. al] σ(mstop~7 TeV) ~10-4 [NLO] FCC-hh analysis meeting FCC-hh analysis meeting

  8. Top reconstruction: Multi-R tagger • The choice of the jet distance parameter (R) is important: • Large enough to contain the topdecay products • But not larger… need to suppress contamination from the underlying event, pile-up and ISR • This leads to large contributions to the jet mass. • Procedure: • Start from an anti-kT SoftDrop PF jetwith R=0.8 • Recluster with smaller R (=0.2 and 0.4) • Calculate jet mass and substructure variables for each case • Expect a very distinct behaviour betweenSignal and Background 0 < pT(gen-top) < 2 TeV 2 < pT(gen-top) < 4 TeV 4 < pT(gen-top) More details in: https://indico.cern.ch/event/697489/contributions/2873783/attachments/1590416/2516620/fcc-hh-susystop-20180126-lg.pdf FCC-hh analysis meeting meeting

  9. Multi-R tagger: performance • Modest gain in efficiency [5-10%] wrt the cut-based approach for the same mistagging rate • As expected in this pT regime • BDT Multi-R inputs: • Softdrop mass, pT, τ32, τ31, τ21 • R = 0.2, 0.4, 0.8 • CC [LHC-based]: • Cut based selection on M(SD) & τ32 • CC-L : 110 < M(SD) < 220 GeV, τ32<0.67 • CC-T : 110 < M(SD) < 220 GeV, τ32<0.57 pT(AK8) < 1 TeV BDT: Multi-R * = CC-T * = CC-L better FCC-hh analysis meeting meeting

  10. Multi-R tagger: performance (2) • Significant improvement in performance for boosted jets • ~ and order of magnitude pT(AK8) > 5 TeV pT(AK8) < 1 TeV Improvement BDT: Multi-R * = CC-T * = CC-L BDT: Multi-R * = CC-T * = CC-L better better However, performance degrades in the case of very boosted tops FCC-hh analysis meeting meeting

  11. Multi-R tagger + tracks • Calorimeter granularity not sufficient for efficient identification of boosted tops [i.e. pT >~4-5 TeV] • ΔR (ECAL) ~ 0.02 , ΔR (HCAL) ~ 0.1 vs. ΔRmax (qi,qj) ~ 0.05 • Exploit tracking – an order of magnitude finer granularity • Procedure [a’la arxiv: 1503.03347 Michele et al.]: • Start from an anti-kT SoftDrop PF jet with R=0.8 • Recluster pfJets with smaller R (=0.2 and 0.4) • Cluster TrkJets with R (=0.2, 0.4, 0.8) • Calculate jet mass and substructure variables for each case CALO-BASED TRK-based FCC-hh analysis meeting meeting

  12. Multi-R + tracks tagger: performance • No improvement from the BDT Multi-R + Tracks tagger ; • Neither from the translated Track – based CC tagger • As expected in this pT regime • BDT Multi-R + tracks inputs: • PFJets and TrkJets with: • Softdrop mass, pT, τ32, τ31, τ21 • R = 0.2, 0.4, 0.8 • CC [LHC-based]: • Cut based selection on M(SD) & τ32 • CC-L : 110 < M(SD) < 220 GeV, τ32<0.67 • CC-T : 110 < M(SD) < 220 GeV, τ32<0.57 • CC-Trk [TrkJets] : • As for CC but selecting on the translatedtrack-based mass and on the track-based subjetiness ratio pT(AK8) < 1 TeV BDT: Multi-R BDT: Multi-R + Trk * = CC-T * = CC-L = CC-Trk-T = CC-Trk-T better FCC-hh analysis meeting meeting

  13. Multi-R + tracks tagger: performance • BDT Multi-R + Tracks tagger: 5-10% gain in efficiency wrt BDT Multi-R • The translated Track – based CC tagger shows significant improvement wrt to the other CC approaches pT(AK8) < 1 TeV pT(AK8) > 5 TeV BDT: Multi-R BDT: Multi-R + Trk * = CC-T * = CC-L = CC-Trk-T = CC-Trk-T BDT: Multi-R BDT: Multi-R + Trk * = CC-T * = CC-L = CC-Trk-T = CC-Trk-T better better FCC-hh analysis meeting meeting

  14. Highlights from search design • Dominant SM processes after baseline selection + top tagging: • ttbar [1Lep] and ttZ (Z->vv) • MET in Z->vv (and ttbar through the “lost” lepton) is aligned with one of the 2 tops • MET in top squarks: more democratically distributed between the 2 tops ttZ(->vv) Top squark FCC-hh analysis meeting

  15. Highlights from search design (2) • Things look as expected :) ttZ(->vv) Top squarks FCC-hh analysis meeting

  16. Highlights from search design (3) • Define an extra handle for stop searches: min Δφ(top1,2;MET) • Powerful variable to suppress ttZ [and ttbar]: • min Δφ(top1,2;MET)> 0.5 : reject ~60% of BKG for ~10% signal ttZ(->nunu) SUSY FCC-hh analysis meeting

  17. Event categorization • On top of the baseline, categorize events based on Nt and Nb: • Fit MET shape: • [for now 7 bins] Nt = 1 Nt >= 2 Nb >= 2 Nb = 1 Nb >= 2 Nb = 1 Top-tagging WP: ~10% mistagging Nt=1, Nb>=2 FCC-hh analysis meeting

  18. Background estimation FCC-hh analysis meeting

  19. Background estimation • Main SM backgrounds after baseline selection: • ttbar [1L] and ttV • Estimation of the main backgrounds: • Measurement of each process in dedicated control samples in “data” • Use MC to translate the measurement into a prediction in the search sample • Rare processes [e.g. ttVV, V, multi-V] from simulation with generous uncertainties FCC-hh analysis meeting

  20. Estimation of the ttbar BKG • Stems from events with leptonically decaying W boson(s) where the lepton is not identified by the vetoes or is out of acceptance : aka “Lost Lepton” (LL) Background • Utilize a 1L control sample [CS] by inverting e/μ veto • “+” : similar kinematics to the search sample [0L] • “-” : limited statistics of the CS • Prediction of LLB : where: • Important: the relative differences [trends] between 0l & 1l need to be similar between data and MC [not the shapes themselves] 1L Selection: - NL = 1 with pT[L] ≥ 30 GeV- MT(L,MET)<100 GeV: suppress potential signal contamination [ TFLL & N(1L) :calculated for each search region] FCC-hh analysis meeting

  21. Estimation of ttV and “Rare” BKGs • The ttZ background is estimated using a multiLepton region: • 3 leptons; OSSF pair consistent with a Z, (pT[Z]>2000 GeV) • Nj>=2, Nb>=2 • Off-Z region used to constrain ttbar background [currently not implemented] • Extract a SF vs pT(Z) • Rare backgrounds [e.g. ttVV, Dibosons,..] are estimated from simulation with 50% uncertainty FCC-hh analysis meeting

  22. Systematic uncertainties • In the ttbar and ttV estimation[particularly in high MET regions] the statistics of the control regions is the dominant uncertainty [~% - 100%]. • 1L and 3L control regions fit simultaneously with the search regions for the signal extraction • Define two scenarios to account for additional sources of systematic uncertainties: • Nominal: 10% [uncertainty uncorrelated across all regions / processes] • Conservative: 20% FCC-hh analysis meeting

  23. Results FCC-hh analysis meeting

  24. Setting the scene • SMS: used in SUSY searches (design & result interpretation) • Minimal set of free parameters to describe a particular set of decay chains • More generic description -> results applicable to other scenarios • But.. there are some simplifications: • eg. full SUSY spectrum not provided; particle properties, (usually) BR=100% for the sparticle decays Set Upper Limitsin cross-section LSP Region kinematically forbidden m[prod. sparticle] < m[LSP] Observed & Expected mass limits based on the nominal production cross-section Excluded parameterspace Produced sparticle Searches for 3rd Generation SUSY

  25. Results @ 3 ab-1 Expected limit Systematics:Nominal scenario [ k-factors applied on both BKG and Signal ] FCC-hh analysis meeting

  26. Results @ 30 ab-1 Expected limit Significance Systematics:Nominal scenario Systematics:Nominal scenario NB: Work on going [smoothing, finer signal scan] FCC-hh analysis meeting

  27. Summary and next steps FCC-hh analysis meeting

  28. Summary • A search for top squarks @ 100 TeV is in place • Focus on all hadronic channel -> largest BR • Tagging ultra-boosted top quarks @ 100 TeV needs detector granularity and improved methods • Multi-R top/W taggers • Track-based substructure variables • Improvements also in the search design • Ttbar & ttV significantly suppressed • For signal models with large mass difference between the top squark and the LSP: • We can reach the mstop~9-10 TeV barrier with 3 ab-1 • We can discover top squarks with mstop~10 TeV with 30 ab-1 5-10x improved background rejectionwrt to cut based approaches FCC-hh analysis meeting

  29. Next steps • For FCC Week: • Small tweaks in top-tagging and search design • Expect ~0.5TeV gain in sensitivity • Exclusion / discovery plots using a scan with higher granularity • Top tagging performance in FullSimulation [?] • Documentation • Longer term plans: • Dedicated search for compressed models • Further improve top-tagging algorithm FCC-hh analysis meeting

  30. Multi-R tagger + tracks + calo (?) • Add also calojets: • … does not improve things .. BDT: Multi-R BDT: Multi-R + Trk BDT: Multi-R + Trk + Calo * = CC-T * = CC-L = CC-Trk-T = CC-Trk-T better FCC-hh analysis meeting meeting

  31. Alternative methods • Similar efficiency with the μ-tagger; a bit better than CC track based • We could include it as input in the BDT to tag the b through non-iso muons • NB. the μ-tagger has very small signal efficiency – by construction Muonic-top tagging Track-based tagging FCC-hh analysis meeting meeting

  32. Top reconstruction: Multi-R tagger (2) • The Signal pT(AK8)<1 TeV Background pT(AK8)<1 TeV ΔR = 0.2 ΔR = 0.4 ΔR = 0.8 ΔR = 0.2 ΔR = 0.4 ΔR = 0.8 Background pT(AK8)>5 TeV Signal pT(AK8)>5 TeV ΔR = 0.2 ΔR = 0.4 ΔR = 0.8 ΔR = 0.2 ΔR = 0.4 ΔR = 0.8 FCC-hh analysis meeting meeting

More Related