Constraints on Higgs width using H*(126) → ZZ events Roberto Covarelli ( University of Rochester ) Seminar at CEA Saclay, 12 May 2014 R. Covarelli

Outline The LHC and the CMS detector After Higgs discovery Theory review of off-shell Higgs production Experimental techniques and combination with on-shell analysis Monte Carlo simulations Analysis of H* → ZZ → 2l2l’ Analysis of H* → ZZ → 2l2n Limits on Higgs width and combination Future plans R. Covarelli

LHC and CMS: operations aa LHC Access Point 1: ATLAS Geneva CERN Meyrin LHC Access Point 5: CMS R. Covarelli 11/15/2018

The CMS detector R. Covarelli 11/15/2018

Looks more and more like the SM Higgs boson After Higgs discovery Great progress since new boson discovery in CMS Observation in boson channels Evidence in fermion channels Mass determination CMS H → ZZ → 4l measurement: 125.6 ± 0.4(stat.) ± 0.2(syst.) GeV Spin/parity studies Angular analyses favor the JPC = 0++ hypothesis arXiv:1401.6527 arXiv:1312.5353 Looks more and more like the SM Higgs boson

Property measurements - width Direct decay width measurements at the peak limited by experimental resolution: f(m) ~ BW(m, G) R(m, s) If G << s, not possible to disentangle natural width SM Higgs width at mH = 125.6 GeV is GH = 4.15 MeV Experimental resolution is s ~ 1-3 GeV for H → ZZ → 4l arXiv:1312.5353 ΓH < 3.4 GeV @ 95% CL

A different idea… Assume a dummy (relativistic BW) resonance “R” with m = 100 and variable width On-shell: Off-shell: Ratio of the two gives G! Experimentally, this never worked before because of tiny off-shell yields and backgrounds “on-shell” region “off-shell” region R. Covarelli

H  WW and ZZ D. De Florian @ Higgs Couplings 2013 R. Covarelli

The idea in detail Off-shell H* → VV (V = W, Z) Peculiar cancellation between BW trend and decay amplitude creates an enhancement of H(126) cross-section at high mVV About 7.6% of total cross-section in the ZZ final state, but can be enhanced by experimental cuts WW ZZ gluon-gluon fusion production H(126) peak Threshold effects at 2mZ and 2mt Recover CPS (~BW) trend N. Kauer and G. Passarino, JHEP 08 (2012) 116 R. Covarelli

Constraint on width F. Caola, K. Melnikov (Phys. Rev. D88 (2013) 054024) J. Campbell et al. (arXiv:1311.3589) Once the “signal strength” m is fixed from an independent source a determination of r is obtained N.B. r-scaling while keeping m fixed is equivalent to coupling scaling Caution: the interference with continuum gg → ZZ is not negligible at high mZZ Can be used to set a constraint on the total Higgs width: R. Covarelli

Monte Carlo simulation gluon-gluon fusion Using latest versions of gg2VV and MCFM (LO in QCD) Including signal H(125.6), background and interference “Running” QCD scales (= mZZ/2) + scale and PDF variations for systematics Signal mZZ-dependent k-factors (NNLO/LO) applied G. Passarino (arXiv:1312.2397) Using results from M. Bonvini et al. (Phys. Rev. D88 (2013) 034032), use kcontinuum = ksignal, assigning an additional 10% uncertainty on this assumption other production modes VBF production is 7% of the total at H(126) peak Slightly enhanced at high mass by trend of sVBF(mZZ) ~ 10% Using PHANTOM to model it, with same settings VH and ttH do not contribute to tail effect R. Covarelli

Analysis procedure Fit r, using one or more variables: P are MC- or data-derived templates for variables in each analysis For a self-contained ZZ analysis use m from CMS on-peak 4-lepton analysis CMS collab. , arXiv:1312.5353: SM width/couplings evaluated at mH = 125.6 GeV Use observed signal strength (“m observed”, ) N.B. An additional assumption we must make is that mggF = mVBF = m (necessary because couplings are in principle different in the two processes, but mVBF not enough constrained by present ZZ data) Expected results are provided also for m = (“m expected”, expected uncertainty from low-mass analysis) R. Covarelli

The 4l and 2l2n final states Generator-level distributions with approximated CMS experimental cuts 4l final state (l = e, m) At high mass, basically only background is qq → ZZ (known at NLO, QCD uncertainties at the level of %) Fully reconstructed state  can use matrix element probabilities of lepton 4-vectors to distinguish between gg and qq production 2l2n final state (l = e, m) Much larger BR (x6) but smaller acceptance (tight pT selection) Rely on transverse mass distributions R. Covarelli N. Kauer and G. Passarino, JHEP 08 (2012) 116 11/15/2018

H → ZZ → 2l2l’

Analysis overview Same event reconstruction and selection as those used in the previous measurement of Higgs boson properties (arXiv: 1312.5353) Event selections: Two pairs of leptons (electrons or muons), isolated and prompt, of opposite sign and same flavor Z1: closest to the Z boson mass Z2: the remaining with highest scalar sum of pT At least one lepton has pT > 20 GeV, and another has pT > 10 GeV 40 < mZ1 < 120 GeV; 12 < mZ2 < 120 GeV Off-shell analysis region: 220 < m4l < 1600 GeV Background: Irreducible background is qq→ZZ, modeled from MC Reducible background is Z+X (Z and WZ, at least one lepton is non-prompt): much smaller, evaluated using a “fake rate” method, with control regions in data

MELA discriminant No changes in selection w.r.t. CMS collab. , arXiv:1312.5353 Lepton pT cuts, Z invariant masses, impact parameter significance, loose isolation In the matrix element likelihood approach (MELA), design a specific discriminant for gg → ZZ production: Built with 7 variables completely describing kinematics (mZ1, mZ2, five angles) Pgg,(qq) are joint probabilities for gg → ZZ, signal + background + interference (qq → ZZ) from MCFM matrix elements R. Covarelli

Input variables to Dgg

m4l and Dgg distributions / yields R. Covarelli

m4l and Dgg distributions / yields R. Covarelli

H → ZZ → 2l2n Missing ET (ETmiss)

Analysis overview Analysis variable is transverse mass: 6 times higher branching fraction compared to 4l final state Branching ratio matters in high mass region where cross section is low No access to Higgs on-shell production  m taken from 4l analysis Z+jets background is several orders of magnitude higher (fake ETmiss due to hadronic energy mis-measurement) Other backgrounds Irreducible: non-resonant ZZ, WZ Non-resonant (not involving a Z boson): top production, WW Analysis variable is transverse mass:

Event selection Z + large ETmiss signature First select a Z → ll: a pair of isolated electrons or muons, pT > 20 GeV, |m(ll) – mZ| < 15 GeV In order to reject WZ: veto 3rd lepton (pT > 10 GeV) In order to reject top processes: veto b-tagged jet or soft muon close to jets (pT > 3 GeV) In order to reject Z+jets: ETmiss > 80 GeV; azimuthal angle between ETmiss and the closest jet: Δϕ > 0.5 To improve sensitivity, selected events are categorized according to number and topology of jets (pT > 30 GeV) VBF / 0 jet / ≥1 jet (but non-VBF) VBF is defined as m(jj) > 500 GeV and |Δη(jj)| > 4

Background estimations qq → ZZ, WZ estimated from MC Non-resonant background (tt, tW, WW) Estimated from data using lepton flavor symmetry: compute the ee/eμ and μμ/eμ ratios in sidebands, and apply the ratios to eμ events in signal region Z+jets background Modeled by photon+jets events in data: reweight photon pT spectrum to match that of dilepton in data, and model ETmiss with photon sample

mT / ET,miss distributions R. Covarelli

mT / ET,miss distributions R. Covarelli

Signal enriched region: ETmiss > 100 GeV and mT > 350 GeV Event yields Signal enriched region: ETmiss > 100 GeV and mT > 350 GeV

Systematic uncertainties (I) Theoretical uncertainties gg → ZZ processes: QCD renormalization and factorization scales varied by a factor of two both up and down, and applied corresponding NNLO K factors; PDF variations by using CT10, MSTW2008 and NNPDF2.1 Additional 10% on continuum gg → ZZ background, accounting for limited knowledge on its NNLO K factor QCD scales and PDF uncertainties on qq → ZZ and WZ backgrounds In the 4l analysis, uncertainty of VBF shapes to account for approximate simulation In the 2l2n analysis, theoretical uncertainties on jet-binning

Systematic uncertainties (II) Experimental uncertainties Lepton trigger, identification, isolation In the 2l2n analysis, uncertainties on lepton momentum scale and jet energy scale are propagated to ETmiss; b-tagging efficiency Background estimations from data Integrated luminosity Limited statistics in MC or data control samples For systematics affect both normalization and shape, variations of shape are taken into account Many are correlated with m measurement  reduced effect on G determination

Results R. Covarelli

4l: results using only m4l or Dgg r < 26.3 (17.0 expected) r < 7.1 (12.7 expected) With m4l observed limit is 1s worse than expected Slight excess of events at high mass Kinematic discriminant strongly suppresses the signal hypothesis for this excess R. Covarelli

4l: 2-dim. result Observed (expected) 95% CL limit: r < 6.6 (11.5) Best fit value: r = 0.5+2.3-0.5 Equivalent to Γ < 27.4 MeV Γ = 2.0+9.6-2.0 MeV

1-dim. fit using mT or ETmiss Results in 2l2ν analysis 1-dim. fit using mT or ETmiss Observed (expected) 95% CL limit: r < 6.4 (10.7) Best fit value: r = 0.2+2.2-0.2 Equivalent to Γ < 26.6 MeV Γ = 0.8+9.1-0.8 MeV ee-only : r < 6.9 (14.3 expected) μμ-only : r < 14.0 (13.7 expected) Counting analysis in “signal-enriched region”: r < 12.4 (16.4 expected)

Combined results Observed (expected) 95% CL limit: r < 4.2 (8.5) p-value = 0.02 Best fit value: r = 0.3+1.5-0.3 Equivalent to Γ < 17.4 (35.3) MeV Γ = 1.4+6.1-1.4 MeV

Conclusions First experimental constraint on Higgs total width using H*(126) → ZZ events has been presented Mild model-dependence Just based on Higgs propagator structure Assumptions on gg → ZZ continuum production beyond LO Assumption of SM production of qq → ZZ and, in general, no other BSM sources enhancing high-mass ZZ yields Combining 4l and 2l2n final states Using variables related to ZZ inv. mass and kinematic discriminants Small deficits in signal regions observed in both channels Combination results: G/GSM < 4.2 (8.5 expected) @ 95% CL  G < 17 MeV (35 MeV expected) @ 95% CL Direct measurements at the peak set a limit of G < 3.4 GeV R. Covarelli

Future plans Addition of 7 TeV CMS data taken in 2011 Implementation of NLO electro-weak corrections for qq → ZZ and WZ backgrounds recently available from several authors: Baglio et al. (arXiv:1307.4331), Bierweiler et al. (arXiv:1305.5402), Gieseke (arXiv:1401.3964) Remove jet binning from 2l2n analysis Recent calculations show that jet bin migration uncertainties may be underestimated Make a joint fit with low-mass 4l analysis Exact correlation of systematic uncertainties R. Covarelli

Back up

4l mass

4l control regions

Input to Dgg in signal-enriched region

4l: limits per final state

Yields vs width (loose Missing ET cut)

Event yields

Systematics

Effect of G / coupling scalings R. Covarelli

PHANTOM settings LO generation Central scale mZZ/√2 NNLO/LO k-factor is 6% and independent on mZZ (from CERN Yellow Report 3) Do not apply explicitly, normalize cross-section at the peak relatively to ggF Central scale mZZ/√2 Same scale and PDF variations as ggF  effect much smaller (1-2%) Signal, background, interference not available separately. Generate total amplitudes with r = 1, 10, 25 (and equal coupling scalings) and extract the 3 components from: R. Covarelli

Full formula of MELA Dgg Depends on parameter a (relative weight of signal in the likelihood ratio). Since the expected exclusion is r ~ 10, use a = 10 R. Covarelli

2l2n: breakdown by channel R. Covarelli