Investigation of the Higgs sector using H->ZZ->4l decays precisely reconstructed by the CMS detector candidate: S. Casasso supervisor: prof. E. Migliore Scuola di Dottorato in "Scienze della Natura e ecnologie Innovative" indirizzo Fisica e Astrofisica
Outline he CMS detector Overview Muon reconstruction in CMS Muon momentum scale and resolution what it is the MuScleFit algorithm momentum calibration: results and validations he H->ZZ (*) ->4l channel overview and candidate's main contributions impact of momentum scale on measurements in the H->ZZ (*) ->4l channel high mass Beyond Standard Model search 2
he CMS detector 3
Muon reconstruction in CMS wo approaches: Outside-in (Global Muons): standalone muons propagated to tracker tracks, if match found -> combined fit Inside-out (racker muons): tracker tracks considered "muons" if matching with standalone tracks / segments Up to p 200-300 GeV: reconstruction is dominated by the racker measurement For p > 300 GeV the Muon System reconstruction improves muon resolution 4
he momentum scale / resolution Scale: bias of the mean value ( 0) affected by any systematic source not averaging to 0 (e.g. misalignment of sensors) Resolution: spread of the distribution around the mean value => affected by any source of smearing in the reconstruction (e.g. hits reconstruction in the tracker, multiple scattering of the particle in the material) 5
he muon momentum scale in CMS he sources of bias in the p measurement depend on the regime: "low p regime" (p < 10 GeV): mis-modeling of material budget in the detector (tracker) "medium-high p regime" (p >~ 10 GeV): residual misalignment of the modules (tracker and muon-chambers) 6
Real life in CMS - 1 In real data the bias on p (or better on curvature) cannot be accessed directly => use well known resonances (Z) to spot residual bias (after alignment of modules) M fit Z is the result of a likelihood fit using a Breit Wigner convoluted with a Crystal Ball plus exponential background Modulation likely due to residual offset of the tracker in the transverse plane Still some "structures" present, especially at high pseudorapidities η 7
Real life in CMS - 2 Data Monte Carlo agreement is one of the main sources of systematics in many analysis in CMS Both scale and resolution do not match perfectly between collision data and simulation "out-of-the-box": MC has up to ~20% better resolution Z peak shifted by up to ~ 0.15% Both scale calibration and extra-smearing on MC are needed for precision physics Arbitrary units/0.08 GeV 0.014 0.012 0.01 0.008 0.006 0.004 0.002 DAA/MC Ratio Before corrections DAA MC fit MZ,data σ CB,data fit MZ,MC σ CB,MC = 91.09 GeV = 1.35 GeV = 91.22 GeV = 1.23 GeV 1.2 0 75 80 85 90 95 100 105 1.15 M µµ (GeV) 1.1 1.05 1 0.95 0.9 0.85 0.8 75 80 85 90 95 100 105 (GeV) M µµ 8
he MuScleFit algorithm he MuScleFit (Muon Scale Fit) algorithm is based on a multiparameter likelihood fit to the invariant mass spectrum of resonances (J/Ψ, Y, Z) decaying to dimuons Goal of the fit is to extract both scale and resolution as analytical functions of muon kinematic variables (p, η, φ) Functional form is "guessed" and parameterized: output of the maximization gives best fit parameters Signal model: NNLO calculation x gaussian resolution Background model: exponential (extracted from calibration samples) 9
he fit strategy "Ansatz function" for scale for the curvature k=q/p : < 0.0015 ~ 0.02 GeV - 1 ~0.00002 GeV - 1 < 0.001 GeV - 1 "Ansatz function" for resolution: j runs over 5 pseudorapidity bins j runs over 12 pseudorapidity bins Use events from selected Z->μμ decays for the calibration 4-step workflow (robust recipe working fine on all the samples): 1. fit resolution => extract resolution before corrections 2. fit scale (p 0 =0) => extract scale using realistic resolution model 3. fit resolution => extract resolution after corrections 4. fit p 0 => extract "global scale factor" for the curvature 10
Extra-smearing on simulation After corrections are applied to both data and simulation, still differences in resolution observed (up to ~10-20%) A gaussian extra-smearing is applied in MC samples to match the resolution observed in data Smearing factor is based on the parameterization of the resolution extracted during the calibration (on the Z) in data (σ data ) and MC (σ MC ) events γ ~ O(1) is an empirical constant 11
Results of the calibration - 1 Arbitrary units/0.08 GeV 0.014 0.012 0.01 0.008 0.006 0.004 After corrections DAA MC fit MZ,data σ CB,data fit MZ,MC σ CB,MC = 91.22 GeV = 1.33 GeV = 91.21 GeV = 1.34 GeV BEFORE calibration 0.002 0 75 80 85 90 95 100 105 1.15 M µµ (GeV) 1.1 1.05 1 0.95 0.9 0.85 0.8 75 80 85 90 95 100 105 M µµ (GeV) DAA/MC Ratio 1.2 Response is flattened (GeV) 92 91.8-1 CMS Preliminary s = 7 ev, L=5.1 fb η <2.4 AFER calibration fit Z M 91.6 Mass shift is recovered 91.4 91.2 91 90.8 90.6 90.4 90.2 2011 data before corrections 2011 data after corrections 90-3 -2-1 0 1 2 3 positive muon φ (rad) 12
Results of the calibration - 2 As a result of the flattening of the bias the overall mass resolution in the data improves by up to ~ 8% (up to ~15% in the endcaps) 13
Data/MC agreement - 1 In order to check the data MC agreement over the whole available phase space, an extensive validation has been set-up: including Z, J/ψ, Υ ->μμ events inspecting relative mass and resolution differences validation binned in the p and η of the muon A lot of effort spent in finding robust and reliable models for signal and background for Z, J/ψ, Υ (both data and simulation) Y1S, Y2S, Y3S Z J/ψ 14
ΔM/M (data-mc) Δσ/σ (data-mc) 0.003 0.002 0.001 0-0.001-0.002-0.003 0.2 0.15 0.1 0.05 0-0.05-0.1-0.15-0.2 Data/MC agreement - 2 BEFORE calibra-ons Z p 20-45 GeV Z p 45-90 GeV Y p 10-20 GeV J/Ψ p 5-7 GeV J/Ψ p 7-10 GeV J/Ψ p 10-15 GeV 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 η BEFORE calibra-ons Z η 0.0-2.4 Y η 0.0-0.7 Y η 0.7-2.4 J/Ψ η 0.0-0.6 J/Ψ η 0.6-1.2 J/Ψ η 1.2-2.0 10 20 30 40 50 60 70 p (GeV) ΔM/M (data-mc) Both resolution and scale systematic deviations are more than halved Δσ/σ (data-mc) Results are striking for the Zs (used in the calibration) but still very good for low mass resonances (i.e. at low p ) too! 0.003 0.002 0.001 0-0.001-0.002-0.003 0.2 0.15 0.1 0.05 0-0.05-0.1-0.15-0.2 AFER calibra-ons Z p 20-45 GeV Z p 45-90 GeV Y p 10-20 GeV J/Ψ p 5-7 GeV J/Ψ p 7-10 GeV J/Ψ p 10-15 GeV 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 η AFER calibra-ons Z η 0.0-2.4 Y η 0.0-0.7 Y η 0.7-2.4 J/Ψ η 0.0-0.6 J/Ψ η 0.6-1.2 J/Ψ η 1.2-2.0 10 20 30 40 50 60 70 p (GeV) 15
Systematics on muon scale o assign a systematic uncertainty on the muon p post-correction we consider 2 use cases (full correlation between the two muons approximation used): Case I: analysis comparing data with a model known from MC Case II: analysis aiming at absolute momentum measurement 16
4 th July 2012 he Higgs (or better "a" Higgs) was discovered by ALAS and CMS at the Large Hadron Collider Both collaborations reached the so-called "5-σ" threshold (thanks to combination of different channels) 17
he Higgs and the "golden channel" he H->ZZ (*) ->4l channel is one of the more powerful: low background rate, clean final state It has been a "riding horse" for the discovery...... it is even more precious now for measurements! December 2011 July 2012 October 2012 March 2013 discovery! (in combination...) discovery! (H->ZZ->4l alone!) 18
he analysis in a nutshell Electron objects Muon objects (ID, reconstruction, efficiency, calibration, smearing,...) Final State Radiation recovery (algorithm, purity estimation) Processing of collisions and simulated (~250) data samples (framework implementation, ntuples production, synchronization of the selection with other groups...) Background estimation (study of sidebands, fake-rate of leptons,...) Events / 3 GeV Event selection (cuts on leptons, cuts on Zs, categorization) 35 30 25 20 15 10 5 0 CMS Data m H =126 GeV Zγ*,ZZ Z+X -1 s = 7 ev, L = 5.1 fb ; 80 100 200 300 400-1 s = 8 ev, L = 19.7 fb 600 m 4l 800 (GeV) Shapes (discriminants, signal shapes low mass, signal shapes high mass,...) Signal efficiency (MC reweighting, interpolation,...) 95% C.L. limit on σ/σ SM 10 1-1 10 CMS -1-1 s = 7 ev, L = 5.1 fb ; s = 8 ev, L = 19.7 fb Observed Expected Expected ± 1σ Expected ± 2σ 100 200 300 400 1000 (GeV) Combined likelihood fit m H 19
Muon momentum scale impact ΔM/M (data - sim.) For muons: MuScleFit momentum calibrations and smearing have been used in the H->ZZ (*) ->4l channel hey have been demonstrated to half both scale and resolution systematics: 0.004 0.003 0.002 0.001 0.000-0.001-0.002-0.003-0.004 CMS muon scale unc. accounts for ~50% of total syst. unc. in the mass measurement in the 4μ channel before calibra-on J/Ψ, p 5-7 GeV J/Ψ, p 7-10 GeV J/Ψ, p 10-15 GeV s = 8 ev, L = 19.7 fb Z, p 20-45 GeV Z, p 45-90 GeV Υ, p 10-20 GeV 0 0.5 1 1.5 2 muon η -1 ΔM/M (data - sim.) 0.003 0.002 0.001 0.000-0.001-0.002-0.003 CMS aaer calibra-on J/Ψ, p 5-7 GeV J/Ψ, p 7-10 GeV J/Ψ, p 10-15 GeV s = 8 ev, L = 19.7 fb Z, p 20-45 GeV Z, p 45-90 GeV Υ, p 10-20 GeV 0 0.5 1 1.5 2 muon η -1 20
In situ validation on Z->4μ In addition to an extensive validation on Z->μμ events: Z->4μ resonant peak ("standard candle") are used to check the scale on data Important cross-check in a region of phase space really close to the newly discovered Higgs boson with ~5 times more events in the peak wrt to a 125 GeV Higgs... No surprise: additional scale is well compatible with 0 (both in data and MC) 21
FSR recovery Final state radiation (FSR) is recovered and added to the 4l system to avoid bias in the mass (with just few events in the peak) Procedure ensures efficient and pure FSR photon selection: p,γ > 2 GeV, η γ < 2.4 ΔR l,γ < 0.5 (*) use Z mass constraint wo-fold gain: 3% better resolution (shrink of mass distribution) + ~2% efficiency (subtract FSR photon from lepton isolation) events CMS Simulation, 500 450 400 350 300 250 200 150 100 50 FSR applied FSR not applied Events affected by FSR M H = 126 GeV s=8 ev 0 80 90 100 110 120 130 140 150 160 [GeV] m 4l+γ Pile-up dependence (*) ΔR = Δφ 2 + Δη 2 22
Signal efficiency and dijet ratio Jet event categorization to test the couplings to fermions and vector bosons: < 2 jets in the event: ~93% from gluon-gluon fusion 2 jets: ~ 50% from VBF, VH, tth Key ingredient to pass from theoretical cross sections to expected yields, include: signal efficiency (with contribution from both categories) di-jet event fraction Both estimated in simulation, after reweighting that accounts for: lepton efficiency measured in data pile-up data-mc differences signal-background interference at high Higgs mass hypothesis Compute separately for each process (ggh, VBF, VH, tth) and final state (4μ, 4e, 2e2μ) 23
Efficiency and dijet ratio interpolation Analytical interpolation between different mass hypothesis corresponding to available MC samples =>smooth description eases the extraction of limits Signal efficiency: Dijet ratio: gluon-gluon fusion efficiency (8 ev) di-jet ratio gluon-gluon fusion (8 ev) di-jet ratio vector boson fusion (8 ev) 24
he legacy paper CMS has decided to publicate for each of the main channels contributing to Higgs measurements (ZZ->4l, WW, γγ, ττ,...) a "legacy paper": should be CMS "legacy" for Run 1 data should include the "state of the art" of both detector performances and analysis techniques H->ZZ->4l legacy analysis is finished and the paper has been submitted to PRD 25
High mass BSM search Data are being re-analyzed in the context of models with a "hidden" scalar field: extra scalar field φ H mixing with SM Higgs (m h < m H ) Single coupling modifier: C'=cos 2 ω rescaling both width and cross section: Adding some extra decay modes for H => BR new : With respect to SM search, need to recompute signal shapes: interference with gg->zz background: huge effect at high masses and strongly dependent on the width 26
Signal shape Signal shape is reweighting to match NNLO prediction via binned K-factors o better describe the high mass shape we use a modified Breit- Wigner multiplied by a universal spline factor (valid for all masses): m H =900 GeV m ZZ 27
Interference parameterization Using gg2vv (3.1.6) generator to compute interference: for each mass hypothesis and C' scale factor Get interference from subtraction: I LO = S+B 2 -S 2 -B 2 I is reweighted to NNLO according to the intermediate recipe: I NNLO = I LO K gg,nnlo Interference pattern has been parameterized as a function of m H hypothesis and C' 28
Signal shapes with interference Interference effect is accounted for by adding it to the signal only shape: S corr (x) = S(x) + r*i(x) heoretical shape is then convoluted with experimental resolution function (double-sided Crystal Ball) m H = 900 GeV, C' 2 = 0.2 C' 2 m H = 900 GeV, C' 2 = 1.0 m H 29
Systematics on the shape heoretical shape systematic uncertainty are evaluated building alternative shapes according to: additive approach: S corr = S NNLO + I LO multiplicative approach: S corr = S NNLO + K NNLO *I LO m H = 800 GeV, C' 2 = 1.0 By varying the amplitude r one can describe the alternative shapes S(x) + r*i(x) Systematic is assigned to r: variation of ~50% covers the envelop of alternative shapes 30
Conclusion and perspectives What I have done in my first 2 years of PhD: contributed to the analysis of data collected in the Run 1 of the LHC by the CMS detector in the channel H->ZZ->4l provided corrections to muon momenta, adopted by the H->4l analysis as well as many other in CMS (e.g. H->μμ) started studying the high mass signal shapes towards the reinterpretation of Run 1 data in the context of Beyond Standard Model searches What I still have to do in my 3 rd (and last year): finalize the inputs and the analysis of the BSM interpretation at high masses contribute to the preparation of the H->ZZ->4l analysis for the Run 2 data starting in 2015 31
Publications Submitted to PRD: "Measurement of the production and decay of a Higgs boson in the four-lepton final state", S. Chatrchyan et al. [CMS Collaboration] 1 : a complete list of my publications can be found at http://inspirehep.net/search?p=exactauthor%3as.casasso.1+ 32
Public and internal notes 33
Conferences 34
Schools and seminars 35
BACKUP 36
OB (racker Outer Barrel) he CMS racker EC (racker Endcap) IB (racker Inner Barrel) ID (racker Inner Disk) 37
Examples of fits - J/ψ 38
Examples of fits Υ 39
Examples of fits Z 40