Top and Higgs Tagging and Where it Helps

Transcription

1 Top and Higgs Tagging and Where it Helps Universität Heidelberg Mainz 2/2

2 Boosted particles at the LHC 994 boosted W 2 jets from heavy Higgs [Seymour] 994 boosted t 3 jets [Seymour] 22 boosted W 2 jets from strongly interacting WW [YSplitter: Butterworth, Cox, Forshaw] 26 boosted t 3 jets from heavy resonances [Agashe, Belyaev, Krupovnickas, Perez, Virzi] 28 boosted H b b [Butterworth, Davison, Rubin, Salam] 28 boosted t 3 jets from heavy resonances [JH tagger: Kaplan, Rehermann, Schwartz, Tweedie] 29 boosted t 3 jets in Higgs production [TP, Salam, Spannowsky] 2 boosted t 3 jets from top partners [: TP, Spannowsky, Takeuchi, Zerwas] 2 boosted t jlν from top partners [: TP, Spannowsky, Takeuchi] 2 first multi-author meta analysis review [BOOST proceedings, Ed: Karagoz, Spannowsky, Vos]

3 Boosted particles at the LHC 994 boosted W 2 jets from heavy Higgs [Seymour] 994 boosted t 3 jets [Seymour] 22 boosted W 2 jets from strongly interacting WW [YSplitter: Butterworth, Cox, Forshaw] 26 boosted t 3 jets from heavy resonances [Agashe, Belyaev, Krupovnickas, Perez, Virzi] 28 boosted H b b [Butterworth, Davison, Rubin, Salam] 28 boosted t 3 jets from heavy resonances [JH tagger: Kaplan, Rehermann, Schwartz, Tweedie] 29 boosted t 3 jets in Higgs production [TP, Salam, Spannowsky] 2 boosted t 3 jets from top partners [: TP, Spannowsky, Takeuchi, Zerwas] 2 boosted t jlν from top partners [: TP, Spannowsky, Takeuchi] 2 first multi-author meta analysis review [BOOST proceedings, Ed: Karagoz, Spannowsky, Vos]

4 Jet Algorithms Definition of jets jet parton duality what are partons in detector? need algorithm to reconstruct what was one parton stable w.r.t inclusion of soft radiation [IR save] crucial for any LHC analysis Different measures [tool: FASTJET; ask Hubert] define jet jet and jet beam distance [and resolution ycut] k T C/A anti-k T y ij = R ij D min `p T,i, p T,j y ij = R ij D y ij = R ij D min p T,i, p T,j () find minimum y min = min kl (y kl, y kb ) (2a) if y min = y kl < y cut combine k and l, go to () (2b) if y min = y kb < y cut remove k, go to () (2c) if y min > y cut, done theoretical and experimental trade-off decisions fat jets: use clustering history y ib = p T,i y ib = y ib = p T,i.

5 Example : VH, H b b New strategy for H bb [Butterworth, Davison, Rubin, Salam] desperately needed [2/3 of all light Higgses; impact Dührssen & SFitter] but killed by continuum Vb b background S: large m bb, boost-dependent R bb B: large m bb only for large R bb S/B: go for large m bb and small R bb, so boost Higgs fat Higgs jet R bb 2m H /p T.8 [like b tag for now] q q V l H b sizeable in boosted regime [p T 3 GeV, few % of total rate] - /σ tot dσ/dp T -2 tth: p T,t -3 tth: p T,H -4 Wjj: p T,j WH: p T,H p T [GeV]

6 Example : VH, H b b New strategy for H bb [Butterworth, Davison, Rubin, Salam] desperately needed [2/3 of all light Higgses; impact Dührssen & SFitter] but killed by continuum Vb b background S: large m bb, boost-dependent R bb B: large m bb only for large R bb S/B: go for large m bb and small R bb, so boost Higgs fat Higgs jet R bb 2m H /p T.8 [like b tag for now] q q V l H b sizeable in boosted regime [p T 3 GeV, few % of total rate] best performance: C/A algorithm jet definition σ S /fb σ B /fb S/ B 3 C/A, R = k, R = SISCone, R =

7 Example : VH, H b b New strategy for H bb [Butterworth, Davison, Rubin, Salam] desperately needed [2/3 of all light Higgses; impact Dührssen & SFitter] but killed by continuum Vb b background S: large m bb, boost-dependent R bb B: large m bb only for large R bb S/B: go for large m bb and small R bb, so boost Higgs fat Higgs jet R bb 2m H /p T.8 [like b tag for now] q q V l H b sizeable in boosted regime [p T 3 GeV, few % of total rate] best performance: C/A algorithm Bottom line, details later combined channels V ll, νν, lν NLO rates [bbv notorious, not from data alone] Z peak as sanity check checked by Freiburg [Piquadio] subjet b tag excellent [7%/%] charm rejection challenging m H ± 8 GeV tough Significance (a) GeV R =.2 Eff = 7% 3GeV R =.7 Eff = 7% 2GeV R =.2 Eff = 6% 3GeV R =.7 Eff = 6% Significance (b) GeV R =.2 Eff = 7% (%) 3GeV R =.7 Eff = 7% (%) 2GeV R =.2 Eff = 6% (2%) 3GeV R =.7 Eff = 6% (2%) b Mistag Probability Higgs Mass (GeV)

8 Example 2: t th, H b b Sad story of t th, H b b [Atlas-Bonn study, CMS-TDR even worse] trigger: t bw + bl + ν reconstruction and rate: t bw bjj continuum background t tb b, t tjj [weighted by b-tag] not a chance: combinatorics: m H in pp 4b tag 2j lν 2 kinematics: peak-on-peak 3 systematics: S/B /9

9 Example 2: t th, H b b Sad story of t th, H b b [Atlas-Bonn study, CMS-TDR even worse] trigger: t bw + bl + ν reconstruction and rate: t bw bjj continuum background t tb b, t tjj [weighted by b-tag] not a chance: combinatorics: m H in pp 4b tag 2j lν 2 kinematics: peak-on-peak 3 systematics: S/B /9 idea [TP, Salam, Spannowsky] pp t l t h H b even harder than VH S/B: R bb <.2; b b pair boosted [solves ] - /σ tot dσ/dp T boosted regime different for S and B [solves 2] see how far we get... [watch S/B for 3] cool: fat Higgs jet + fat top jet -2-3 tth: p T,t tth: p T,H -4 Wjj: p T,j WH: p T,H p T [GeV]

10 Example 2: t th, H b b Sad story of t th, H b b [Atlas-Bonn study, CMS-TDR even worse] trigger: t bw + bl + ν reconstruction and rate: t bw bjj continuum background t tb b, t tjj [weighted by b-tag] not a chance: combinatorics: m H in pp 4b tag 2j lν 2 kinematics: peak-on-peak 3 systematics: S/B /9 Another bottom line require tagged top and Higgs trigger on lepton remove Higgs as t l b plus QCD 3rd b tag in continuum [costing S/ B] only continuum t tb b left per fb signal t tz t tb b t t+jets events after acceptance events with one top tag events with m bb = 3 GeV corresponding to subjet pairings subjet pairings two b tags including a third b tag

11 Example 2: t th, H b b Sad story of t th, H b b [Atlas-Bonn study, CMS-TDR even worse] trigger: t bw + bl + ν reconstruction and rate: t bw bjj continuum background t tb b, t tjj [weighted by b-tag] not a chance: combinatorics: m H in pp 4b tag 2j lν 2 kinematics: peak-on-peak 3 systematics: S/B /9 Another bottom line require tagged top and Higgs trigger on lepton remove Higgs as t l b plus QCD 3rd b tag in continuum [costing S/ B] only continuum t tb b left m H S S/B S/ B fb 5 57 / (5.7) 2 48 / (5.) 3 29 / (3.).6 dσ/dm b b dσ/dm b b [fb/5 GeV] [fb/5 GeV] 9 2 t th t tz t tjj t tb b t th t tz t tjj t tb b 5 m b b 8 [GeV]

12 Example 3: H b b in SUSY cascades Higgs in cascade decays [Kribs, Martin, Roy, Spannowsky] idea: find Higgs in cascade decays [Cambridge] BSM sample after missing energy or hard γ cut blind Higgs tag over remaining event [QCD rejection?] side bin analysis in m bb more to follow... - bb invariant mass, L = fb, Number of Events/6. GeV 6 tt + jets s = 4 TeV W + bb Z + bb W + jets Z + jets SUSY M bb

13 Higgs tag for busy QCD environment [BDRS; TP, Salam, Spannowsky] uncluster one-by-one: j j + j 2 unbalanced m j >.8m j means QCD; discard j 2 2 soft m j < 3 GeV means QCD; keep j double b tag [possibly add balance criterion] three leading J = p T, p T,2 ( R 2 ) 4 vs m bb no mass constraint side bin typical mis-tag probability < 5 underlying event and pileup deadly filter reconstruction jets [Butterworth Salam, cf pruning, trimming] zoomed-in C/A analysis with R filt = min(.3, R bb /2) reconstruct m H w/ one QCD jet

14 Higgs tag for busy QCD environment [BDRS; TP, Salam, Spannowsky] uncluster one-by-one: j j + j 2 unbalanced m j >.8m j means QCD; discard j 2 2 soft m j < 3 GeV means QCD; keep j double b tag [possibly add balance criterion] three leading J = p T, p T,2 ( R 2 ) 4 vs m bb no mass constraint side bin typical mis-tag probability < 5 underlying event and pileup deadly filter reconstruction jets [Butterworth Salam, cf pruning, trimming] zoomed-in C/A analysis with R filt = min(.3, R bb /2) reconstruct m H w/ one QCD jet Better than traditional b jets no combinatorial choices more soft partons included in m H b tagging easier than in continuum QCD features useful [Soper & Spannowsky]...

15 Top tagger Highly boosted top quarks identify hadronic tops with p T 8 GeV isolation and b tagging challenging [Kaplan, Rehermann, Schwartz, Tweedie; Princeton, Seattle...] C/A algorithm with p T drop criterion [Hopkins tagger, no b tag] top mass included, no sidebins ATLAS studies for semileptonic top pairs [adapted Y-splitter, full sim, ATLAS-2-8]

16 Top tagger Highly boosted top quarks identify hadronic tops with p T 8 GeV isolation and b tagging challenging [Kaplan, Rehermann, Schwartz, Tweedie; Princeton, Seattle...] C/A algorithm with p T drop criterion [Hopkins tagger, no b tag] top mass included, no sidebins ATLAS studies for semileptonic top pairs [adapted Y-splitter, full sim, ATLAS-2-8] Out first top tag [TP, Salam, Spannowsky, Takeuchi] start with C/A jet [R =.5] [Johns Hopkins] uncluster one-by-one: j j + j 2 unbalanced m j >.8m j means QCD; discard j 2 2 soft m j < 3 GeV means QCD; keep j top decay kinematics in relevant substructures.2 reconstruct m W = GeV. reconstruct m t = GeV helicity angle cos θ t,j >.7.8 no b tag needed.6 Events normalized to unity.4.2 top jets gluon jets quark jets cosθ h

17 Top tagger Highly boosted top quarks identify hadronic tops with p T 8 GeV isolation and b tagging challenging [Kaplan, Rehermann, Schwartz, Tweedie; Princeton, Seattle...] C/A algorithm with p T drop criterion [Hopkins tagger, no b tag] top mass included, no sidebins ATLAS studies for semileptonic top pairs [adapted Y-splitter, full sim, ATLAS-2-8] Out first top tag [TP, Salam, Spannowsky, Takeuchi] start with C/A jet [R =.5] [Johns Hopkins] uncluster one-by-one: j j + j 2 unbalanced m j >.8m j means QCD; discard j 2 2 soft m j < 3 GeV means QCD; keep j.25 top decay kinematics in relevant substructures reconstruct m W = GeV.2 reconstruct m t = GeV.5 helicity angle cos θ t,j >.7 no b tag needed. filtering w/ 2 QCD jets.5 t th t tb b t tz m rec W m rec t

18 Applicable: [TP, Salam, Spannowsky, Takeuchi] extend lower p T 25 GeV testable in Standard Model t t events Number of tops QCD top tagged top W+jets p [GeV] T

19 Applicable: [TP, Salam, Spannowsky, Takeuchi] extend lower p T 25 GeV testable in Standard Model t t events new kinematic selection: m jjj, m (), m (2) jj jj [no boost] m 23 /m 23 m2 =m W.8 m 3 =mw m 23 /m 23 m2 =m W.8 m 3 =mw m 23 /m 23 m2 =m W.8 m 3 =mw m 23 =m W m 23 =m W m 23 =m W arctan m 3 /m arctan m 3 /m arctan m 3 /m 2

20 Applicable: [TP, Salam, Spannowsky, Takeuchi] extend lower p T 25 GeV testable in Standard Model t t events new kinematic selection: m jjj, m (), m (2) jj jj [no boost] R bjj P T [GeV] R bjj P T [GeV]

21 Applicable: [TP, Salam, Spannowsky, Takeuchi] extend lower p T 25 GeV testable in Standard Model t t events new kinematic selection: m jjj, m (), m (2) jj jj momentum reconstruction for free [no boost] Number of tops rec R(p part,p ) Number of tops rec p/p

22 Applicable: [TP, Salam, Spannowsky, Takeuchi] extend lower p T 25 GeV testable in Standard Model t t events new kinematic selection: m jjj, m (), m (2) jj jj momentum reconstruction for free use color activity [Baryakthar, Hook, Janowiak, Wacker] tested by ATLAS Heidelberg [Kasieczka & Schätzel] hadronic top like tagged b [no boost] Efficiencies inside fat jet tagged tagged (unmatched) p [GeV] T t t QCD W +jets p T min [ GeV],t 2 3 one fat jet % % % % % two fat jets 44% 57% 7% 53% 5% rel to one fat jet one top tag 23% 37% 5% 2.% 3.9% rel to one fat jet two top tags 2.% 4.5% 8.5%.27%.7% rel to one fat jet 4.5% 8.% 2%.5%.5% rel to two fat jets

23 Stops pairs as the first application [TP, Spannowsky, Takeuchi, Zerwas] stop crucial for hierarchy problem [review: Morrissey, TP, Tait] comparison to other top partners [Meade & Reece] dark matter means difficult semi-leptonic channel [possibly impossible] hadronic: t t t χ t χ [CMS: leptons as spontaneous life guards; Meade & Reece overly optimistic] t t events in fb t t t t QCD W +jets Z +jets S/B S/ B fb [ GeV] m t p T,j > 2 GeV, l veto n/a 3. 5 /E T > 5 GeV first top tag second top tag b tag m T 2 > 25 GeV

24 Stops pairs as the first application [TP, Spannowsky, Takeuchi, Zerwas] stop crucial for hierarchy problem [review: Morrissey, TP, Tait] comparison to other top partners [Meade & Reece] dark matter means difficult semi-leptonic channel [possibly impossible] hadronic: t t t χ t χ stop mass from m T 2 endpoint [like sleptons or sbottoms] m T 2 ( ˆm χ) = [CMS: leptons as spontaneous life guards; Meade & Reece overly optimistic]»! min max m T,j (q j ; ˆm χ) < m t /p T =q +q j 2 t t hadronic search as easy as b b + /E T [/2 GeV] d dm σ T2 σ.4 test MLSP =98 GeV tt M stop =34 GeV.2 M stop =54 GeV M T2 [GeV]

25 Leptonic top tag [Thaler & Wang; Rehermann & Tweedie; TP, Spannowsky, Takeuchi] known: masses of top decay products unknown: 3-momentum of neutrino measured: E b, E l, m bl [rest frame] W and t mass constraints third parameter elsewhere do not use measured /p T vector

26 Leptonic top tag [Thaler & Wang; Rehermann & Tweedie; TP, Spannowsky, Takeuchi] known: masses of top decay products unknown: 3-momentum of neutrino measured: E b, E l, m bl [rest frame] W and t mass constraints third parameter elsewhere do not use measured /p T vector neutrino coordinates leading in b l direction sub-leading in b l decay plane [orthogonal approx p ν = ] sub-leading orthogonal to decay plane [decay plance approx p ν = ] components (p ν, p ν ) x p = 2-3 GeV t x p = 4-5 GeV t x p = 6-7 GeV t x p = 8-9 GeV t x x x x

27 Leptonic top tag [Thaler & Wang; Rehermann & Tweedie; TP, Spannowsky, Takeuchi] known: masses of top decay products unknown: 3-momentum of neutrino measured: E b, E l, m bl [rest frame] W and t mass constraints third parameter elsewhere do not use measured /p T vector neutrino coordinates leading in b l direction sub-leading in b l decay plane [orthogonal approx p ν = ] sub-leading orthogonal to decay plane [decay plance approx p ν = ] semileptonic top partners at LHC: At the LHC, combinatorics make it unlikely that we will be able to observe stop pair production with a decay to a semileptonic top pair and missing energy. [TP, Spannowsky, Takeuchi, Zerwas]

28 Leptonic top tag [Thaler & Wang; Rehermann & Tweedie; TP, Spannowsky, Takeuchi] known: masses of top decay products unknown: 3-momentum of neutrino measured: E b, E l, m bl [rest frame] W and t mass constraints third parameter elsewhere do not use measured /p T vector neutrino coordinates leading in b l direction sub-leading in b l decay plane [orthogonal approx p ν = ] sub-leading orthogonal to decay plane [decay plance approx p ν = ] semileptonic top partners at LHC: At the LHC, combinatorics make it unlikely that we will be able to observe stop pair production with a decay to a semileptonic top pair and missing energy. [TP, Spannowsky, Takeuchi, Zerwas] wrong!

29 Leptonic top tag [Thaler & Wang; Rehermann & Tweedie; TP, Spannowsky, Takeuchi] known: masses of top decay products unknown: 3-momentum of neutrino measured: E b, E l, m bl [rest frame] W and t mass constraints third parameter elsewhere do not use measured /p T vector neutrino coordinates leading in b l direction sub-leading in b l decay plane [orthogonal approx p ν = ] sub-leading orthogonal to decay plane [decay plance approx p ν = ] semileptonic top partners at LHC: use approximate Φ( /p T, ˆp t ) dσ d φ tot σ.3.2 p ν φ π π/2. -π/2 - π -π 2 π 2 π φ -π p/ [GeV] T

30 Leptonic top tag [Thaler & Wang; Rehermann & Tweedie; TP, Spannowsky, Takeuchi] known: masses of top decay products unknown: 3-momentum of neutrino measured: E b, E l, m bl [rest frame] W and t mass constraints third parameter elsewhere do not use measured /p T vector neutrino coordinates leading in b l direction sub-leading in b l decay plane [orthogonal approx p ν = ] sub-leading orthogonal to decay plane [decay plance approx p ν = ] semileptonic top partners at LHC: use approximate Φ( /p T, ˆp t ) top partner decays observable orthogonal approximation decay plane approximation t t t t W +jets S/B t t t t W+jets S/B [ GeV] m t.-5. base cuts approximation p T est > 2GeV /p T vs. φ cut

31 Outlook Aspirin of LHC phenomenology VH: bringing back 2/3 of light Higgses t th: curing combinatorics and backgrounds SUSY cascades: curing lack of analysis idea... Z etc: improving mass resolution t t : curing backgrounds... code as FASTJET add-on [ plehn/] implemented and tested by ATLAS, improvements welcome leptonic tagger soon public LHC lecture notes arxiv:9.482 BOOST review arxiv:2.542

32