Identifying top quarks in the pursuit of new physics Siddharth - PowerPoint PPT Presentation

Identifying top quarks in the pursuit of new physics Siddharth Narayanan, MIT PPC MIT LNS Seminar, 20/02/2018 S. Narayanan (MIT) LNS Seminar 20/02/2018 1 / 42

Searching for new physics at the LHC ◮ The LHC is a particle factory ◮ Many ways to search for beyond-the-Standard Model physics at the LHC: ◮ Produce SM particles and precisely measure their properties ◮ Produce SM bound states and study interesting decay channels ◮ Produce BSM particles and identify them ◮ . . . ◮ Identify BSM particles by looking for: ◮ Resonant final states ◮ Exotic decays (e.g. semi-stable particles) ◮ Particles with small couplings to SM particles ◮ . . . S. Narayanan (MIT) LNS Seminar 20/02/2018 2 / 42

Seeing the invisible ◮ Particles that do not interact with our detector are “invisible” for practical purposes ◮ Their presence must be inferred through momentum conservation: ◮ This momentum imbalance is referred to as p miss T ◮ Invisible particles include dark matter candidates L’essentiel est invisible pour les yeux. - A. Exup´ ery S. Narayanan (MIT) LNS Seminar 20/02/2018 3 / 42

Hadronic top quarks and missing momentum χ u ◮ Choice of SM particle probes different models and V parameter spaces ¯ χ u ◮ Light quarks or gluons ◮ W/Z/H/γ bosons ◮ Top quarks g t ◮ Single top quark + p miss implies flavor-violating new ¯ T ψ d physics ◮ Can have implications for baryogenesis and DM ◮ Require the top quark to decay hadronically: φ ◮ Larger branching ratio ◮ p miss is purely due to BSM particlecs T s ¯ t S. Narayanan (MIT) LNS Seminar 20/02/2018 4 / 42

Outline ◮ The mono-top search at CMS ◮ Constructing principle observables ◮ Identifying hadronic top quarks ◮ Constraining backgrounds ◮ Interpreting results ◮ New approaches to top-tagging S. Narayanan (MIT) LNS Seminar 20/02/2018 5 / 42

Search for top+ p miss T S. Narayanan (MIT) LNS Seminar 20/02/2018 6 / 42

Compact Muon Solenoid ◮ General-purpose detector ◮ p miss is constructed as: T   � p miss � = − � p i T   i ∈ particles T where the sum is over all identifiable particles S. Narayanan (MIT) LNS Seminar 20/02/2018 7 / 42

Compact Muon Solenoid Solenoid ◮ 3 . 8 T field parallel to beam Silicon tracker ◮ Inner layers of pixels and outer layers of strip detectors ◮ Momentum measurement of charged hadrons, electrons, muons ◮ Track vertexing to ID ◮ . . . pile-up noise ◮ . . . B -meson decays S. Narayanan (MIT) LNS Seminar 20/02/2018 8 / 42

Compact Muon Solenoid Calorimeters ◮ EM: homogenous, PbWO 4 crystals ◮ Hadronic: sampling, brass and plastic scintillator ◮ Energy resolution and large coverage of calorimeters critical to p miss resolution T Muon chambers ◮ Various ionization detectors ◮ Used to ID muons and improve momentum measurement S. Narayanan (MIT) LNS Seminar 20/02/2018 9 / 42

Observables of mono-top final state ◮ Large p miss and hadronic top decay define mono-top T ◮ Top quark decays to 3 quarks ⇒ 3 jets ◮ “Jet” is algorithm-dependent ◮ e.g. anti- k T algorithm with radius of R = 0 . 4 in η - φ plane ◮ Signal models produce more energetic jets than background processes ◮ Separation between daughter quarks scales as ∆ R ∼ 2 m t /p T ◮ If we want to look for highly-boosted top quarks, we cannot resolve decay products as separate jets S. Narayanan (MIT) LNS Seminar 20/02/2018 10 / 42

Reconstruction of top quark ◮ Replace 3 AK 0.4 jets (AK4) with a single CA 1.5 jet (CA15) ◮ 0 . 4 → 1 . 5: much wider radius ◮ Anti- k T → Cambridge-Aachen: more geometrical clustering ◮ Large radius allows single algorithm to reconstruct range of top quark momenta ◮ As low as p T ∼ 250 GeV ◮ These are big jets ◮ R = 1 . 5 can contain up to half the detector ◮ Unwanted particles can sneak into top jet ◮ Fake top jets from combinatorial q/g - S. Narayanan (MIT) LNS Seminar 20/02/2018 11 / 42

Identifying top jets ◮ Remove pile-up contamination from event ◮ Pile-Up Per Particle Identification (PUPPI) algorithm [ arXiv:1407.6013 ] ◮ Likelihood given particle is from primary vertex ◮ Remove soft and wide-angle radiation from jet ◮ Soft drop grooming [ arXiv:1402.2657 ] ◮ Used to improve mass resolution and define “subjets” ◮ Identify b subjets ◮ Probability based on signatures of B meson decays, including displaced decay vertex 35.8 fb -1 (13 TeV) -1 36 fb , s = 13 TeV, 2016 a.u. 13 Jets / 0.02 10 CMS Preliminary Data Data CMS 0.16 10 12 Data t t t t 11 W+jets W+jets 10 Preliminary b 0.14 Z+jets Z+jets 10 10 Single t Single t Muon Enriched Multijet sample 9 c 10 0.12 Diboson Diboson Soft drop subjets of Muon-tagged AK8 jets 8 10 QCD QCD udsg p (AK8 jets) > 350 GeV 7 0.1 10 T 6 10 0.08 5 10 10 4 0.06 3 10 10 2 0.04 10 0.02 1 0 Data/MC Data-Exp 1.5 0.4 0 50 100 150 200 250 Exp 0.2 0 1 − 0.2 − 0.4 0.5 0 50 100 150 200 250 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 fatjet m [GeV] SD CSVv2 Discriminator S. Narayanan (MIT) LNS Seminar 20/02/2018 12 / 42

Jet substructure ◮ A top jet is expected to have a structure consistent with 3 partons ◮ Jets from light quarks typically do not have distinct prongs ◮ Observables that are sensitive to such structure are referred to as substructure S. Narayanan (MIT) LNS Seminar 20/02/2018 13 / 42

Substructure variables ◮ N -subjettiness [ arXiv:1011.2268 ] ◮ τ N are measure of compatibility of jet with N -axis hypothesis ◮ HEPTopTagger [ arXiv:1312.1504 ] ◮ Decluster the jet into subjets and re-combine them to reconstruct W and t ◮ Variable of interest is f rec ∼ m W /m t ◮ Energy correlation functions [ arXiv:1609.07473 ] ◮ Defines variables e ( α, N, a ) sensitive to N -point correlations in the jet 0.22 Events/0.02 Events 0.16 Events/ 0.09 CMS Preliminary CMS Preliminary CMS Preliminary QCD QCD 0.2 QCD QCD QCD QCD 0.14 0.08 110 < m < 210 GeV 0.18 110 < m < 210 GeV 110 < m < 210 GeV SD SD SD Top Top Top Top Top Top 0.12 0.07 0.16 0.14 0.06 0.1 0.12 0.05 0.08 0.1 0.04 0.06 0.08 0.03 0.06 0.04 0.02 0.04 0.02 0.01 0.02 0 0 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 τ 2 Groomed HTT f e(2,4,2)/e(1,3,2) 32 rec S. Narayanan (MIT) LNS Seminar 20/02/2018 14 / 42

ECF ratios ◮ Expect a top jet to have strong 3-point correlations, but not 4-point correlations ◮ e ( N = 4) /e ( N = 3) ∼ 0 ◮ Both N = 3 and N = 4 should be weak for q/g jets ◮ e ( N = 4) /e ( N = 3) > 0 ◮ Ratio proposed in arXiv:1609.07473 is e (2 , 4 , β ) N ( β ) = 3 ( e (1 , 3 , β )) 2 ◮ Can naturally extend this to a wider class of dimensionless variables: ψ ( a, α, N ; b, β, M ) ≡ e ( a, N, α ) e ( b, M, β ) x , where M ≤ N and x = aα bβ S. Narayanan (MIT) LNS Seminar 20/02/2018 15 / 42

Non-trivial ECF ratios ◮ Turns out many correlation function ratios can separate signal and background ◮ Of course, most combinations are highly correlated or not useful Events Events 0.09 Events CMS Preliminary CMS Preliminary CMS Preliminary 0.25 QCD QCD QCD QCD QCD QCD 0.14 0.08 110 < m < 210 GeV 110 < m < 210 GeV 110 < m < 210 GeV SD SD SD Top Top Top Top Top Top 0.12 0.07 0.2 0.06 0.1 0.15 0.05 0.08 0.04 0.06 0.1 0.03 0.04 0.02 0.05 0.02 0.01 0 0 0 2 3 4 5 6 7 8 9 10 0 0.05 0.1 0.15 0.2 0.25 0 0.5 1 1.5 2 2.5 2 1/2 2 e(1,2,2)/e(1,2,1) e(3,3,2)/e(3,3,4) e(1,4,4)/e(1,3,2) e ( N = 2) /e ( N = 2) e ( N = 3) /e ( N = 3) e ( N = 4) /e ( N = 3) S. Narayanan (MIT) LNS Seminar 20/02/2018 16 / 42

Building a multivariate tagger 1 background acceptance ◮ ECFs and τ N are computed after soft CMS Preliminary drop grooming is applied 110 < m < 210 GeV SD ◮ Space of ECF ratios is pruned based on 2 criteria: ◮ Separation power − 1 10 ◮ Agreement between MC simulation τ SD 11 ECF + + f BDT 32 rec and data 11 ECF BDT ◮ Use a boosted decision tree on 11 ECF 50 ECF BDT ratios, f rec , τ 3 /τ 2 τ SD ǫ bkg ( ǫ sig = 0 . 5) +f BDT 32 rec − 2 10 ECF+ τ 32 + f rec 4 . 7% τ Groomed 32 τ 32 + f rec 6 . 2% τ 32 6 . 9% 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 signal efficiency S. Narayanan (MIT) LNS Seminar 20/02/2018 17 / 42