Top content in ATLAS ttH() measurements Jennet Dickinson for the - PowerPoint PPT Presentation

Top content in ATLAS ttH(ɣɣ) measurements Jennet Dickinson for the ATLAS Collaboration Moriond EW March 17, 2019

ttH(ɣɣ) analysis strategy 1 Fraction of Events • Require 2 photons passing Cont. Bkg. 0.9 ATLAS Preliminary NTI Control Region -1 s = 13 TeV, 139 fb t t H 0.8 Non-t t H Higgs Had region tight ID and isolation criteria 0.7 0.35 0.6 0.3 • Separate events by decay 0.5 0.25 0.2 0.4 0.15 of top quarks 0.1 0.3 0.05 0.2 0.9 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1 0.1 (1) Hadronic region (4 categories) 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 (2) Leptonic region (3 categories) BDT Output • Boosted Decision Trees (BDTs) are trained on object-level variables to separate ttH MC signal from background – Background is modeled by data control sample failing photon tight ID or isolation (NTI) 3/17/19 Jennet Dickinson 2

ttH(ɣɣ) signal + background fit • Signal strength is extracted from a maximum likelihood fit to all categories • Determination of continuum background is completely data-driven (normalization and shape) • How top-like is this Sum of Weights / 1.375 GeV background? ATLAS Preliminary Data 30 -1 s = 13 TeV, 139 fb Continuum Background Total Background m = 125.09 GeV – Primarily composed of ttɣɣ H 25 Signal + Background All categories ln(1+S/B) weighted sum and ɣɣ + jets 20 15 – Small contribution from jets 10 faking photons 5 • Study this background using 0 110 120 130 140 150 160 m [GeV] reconstructed hadronic tops g g 3/17/19 Jennet Dickinson 3

Reconstructing hadronic tops q • Hadronic top decays correspond to 3 quarks ~ 3 jets in ttH W q 0 t • Goal: identify these 3 jets – Many possible combinations! b • Train a dedicated BDT for top reconstruction – Signal: jet triplets truth-matched to tops (ttH MC) – Background: other triplets (ttH MC) – Training variables: momenta & b-tag score of jets, angles between jets, m jjj • The jet triplet in each event with highest BDT score is designated as the top candidate 3/17/19 Jennet Dickinson 4

Template fit method • Exploit the shape difference in the top candidate mass between samples with/without true tops • Construct templates from top mass distributions in ttɣɣ, ɣɣ+jets and ttH Monte Carlo • Decompose the continuum Events ATLAS Preliminary g g + jets background by performing a -1 g g 100 s = 13 TeV, 139 fb t t t t H Fitted total 80 template fit to data: Data 60 Two Tightest Had Categories af tt γγ ( m ) + bf γγ ( m ) + n ttH 40 SM f ttH ( m ) 20 n data 0 0 50 100 150 200 250 300 350 400 450 500 Top candidate mass [GeV] 3/17/19 Jennet Dickinson 5

Top fractions in the hadronic region • Tighter ttH(ɣɣ) selection should give more top-like background ttɣɣ fraction (looser region) ttɣɣ fraction (tighter region) a = 0 . 31 ± 0 . 17 a = 0 . 21 ± 0 . 06 Events 600 Events ATLAS Preliminary ATLAS Preliminary g g g g + jets + jets -1 -1 g g g g s = 13 TeV, 139 fb 100 s = 13 TeV, 139 fb t t t t 500 t t H t t H Fitted total Fitted total 80 400 Data Data 60 300 All Had Categories Two Tightest Had Categories 40 200 20 100 0 0 0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500 Top candidate mass [GeV] Top candidate mass [GeV] • These estimates of background passing ttH(ɣɣ) selection direct further optimization efforts 3/17/19 Jennet Dickinson 6

7

Thank you! 8

Backup 3/17/19 Jennet Dickinson 9

References • ATLAS publications – ttH discovery: Phys. Lett. B 784 (2018) 173 – Latest ttH(ɣɣ): ANA-HIGG-2018-59-CONF • Other – LHC HIGGS XS WG 3/17/19 Jennet Dickinson 10

ttH production in pp collisions • ttH production is a direct probe of the Higgs-top Yukawa coupling • Measurements of this process are challenging – Low rate: at 13 TeV, SM σ ttH = 507 fb – Complex final states: LHC HIGGS XS WG decay products of 2 tops and Higgs 3/17/19 Jennet Dickinson 11

ttH(ɣɣ) Analysis strategy • Events are pre-selected in two groups: (1) leptonic (≥1 b-jet, ≥1 leptons) – 3 BDT categories (2) hadronic (≥1 b-jet, ≥3 jets, 0 leptons) – 4 BDT categories • Events are then further divided into categories based on an XGBoost BDT discriminant – Training uses energy and direction of photons, jets, leptons, jet b-tag flag, MET and MET_φ 3/17/19 Jennet Dickinson 12

ttH(ɣɣ) category definition in the hadronic channel • Define four hadronic ttH categories with different S/B by slicing in BDT score – Reject events with BDT score < 0.91 • Tight BDT categories 1 Fraction of Events Cont. Bkg. 0.9 ATLAS Preliminary NTI Control Region have lower statistics, but -1 s = 13 TeV, 139 fb t t H 0.8 Non-t t H Higgs Had region 0.7 higher ttH purity and 0.35 0.6 0.3 0.5 0.25 better S/B ratio 0.2 0.4 0.15 0.3 0.1 0.05 – These are the most 0.2 0.9 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1 0.1 powerful categories 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 BDT Output 3/17/19 Jennet Dickinson 13

ttH(ɣɣ) category definition in the leptonic channel • Define three leptonic ttH categories with different S/B by slicing in BDT score – Reject events with BDT score < 0.70 • Again, tightest BDT 1 Fraction of Events Cont. Bkg. 0.9 ATLAS Preliminary NTI Control Region category is the most -1 s = 13 TeV, 139 fb t t H 0.8 Non-t t H Higgs Lep region 0.7 powerful due to high S/B 0.8 0.6 0.7 0.5 0.6 0.5 • Statistics in the leptonic 0.4 0.4 0.3 0.3 0.2 channel are lower 0.2 0.1 0.7 0.75 0.8 0.85 0.9 0.95 1 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 BDT Output 3/17/19 Jennet Dickinson 14