1. 陶军全 中科院高能所 Junquan Tao (IHEP/CAS, Beijing)
Measurements of Higgs boson production in the diphoton decay channel at √s =13 TeV
陶军全 中科院高能所
Junquan Tao (IHEP/CAS, Beijing)
中国物理学会高能物理分会第十二届全国粒子物理学术会议
2016年8月22-26日 安徽合肥

2. H→ at CMS: reminder
Higgs discovery on 4th July 2012 : most sensitive channel
Run1 legacy result with fb-1 at 7TeV and fb-1 at 8TeV data : observation of diphoton decay mode
Phy. Let. B 716 (2012) 30–61
CMS public the first results with TeV 2.7 fb-1 data on Moriond2016: ~2σ at GeV
In this talk, update measurements of H→ at 13TeV with fb-1 data : significance, signal strength, couplings, fiducial XS
Eur. Phys. J. C (2014) 74:3076
~124.7 GeV 5.7σ(obs.)/5.2σ(exp.)

3. Outline Introduction Analysis strategy and analysis elements
Event categorization, Signal and Background Modelling
Measurement results : significance, signal strength and couplings
Fiducial XS measurement
Summary

4. Introduction Clean final state with two highly energetic photons
Final state fully reconstructed with high resolution
Good mass resolution : ~ 1-2% m
Very small branching fraction (~0.2%)
Large backgrounds (, j, jj)
Search for a narrow peak on a falling
background in mass distribution
Exclusive categories targeting:
gluon-gluon fusion (ggH),
vector boson fusion (VBF)
and ttH production modes

5. Analysis strategy and analysis elements

6. Analysis strategy
Photons energy reconstructed and corrected, rejection of fake photons (photon id)
Events categorized into classes (production mechanism, mass resolution, S/B) to improve the analysis sensitivity
Extraction of signal through fit of di-photon invariant mass spectrum in each event class

7. Photon energy Electro-magnetic calorimeter response
Energy and its uncertainty corrected for local and global shower containment: regression targeting Etrue/Ereco
Scale vs time and resolution calibration: Z→ee peak used as reference
Corrected energies and resolutions used in the analysis
• corrected for change in time
• inter-calibrated to be uniform in η/φ
• adjustment of absolute scale

8. Vertex identification
Vertex assignment correct within 1 cm → negligible impact on mass resolution
No ionization in the tracker for photons
Multi-variate approach for vertex identification
Second MVA estimates probability of correct vertex choice, used for di-photon classification
Method validated on Z→μμ (+j for converted ) events, where vertex found after removing muon tracks
kinematic correlations and track distribution imbalance
Conversion information

9. Event and photon selection
Double-photon trigger selection based on transverse energy, m
, isolation and electromagnetic shower shapes variables
Minimal pre-selection, similar to but tighter than trigger selections
Efficiency is measured in MC with SFs from data and simulation with tag & probe method
• Z→ee events used for all cuts other than electron veto
• Z→μμ events for electron veto

10. Photon identification
MVA based photon ID classifier (BDT) to discriminate between prompt and fake photons
Inputs and output of the MVA are validated on data and MC in Z→ee and Z→μμ events
Two photon BDT scores are used as inputs of diphoton BDT after a looser direct cut at > -0.9

11. Diphoton BDT
Multivariate discriminator (BDT) used to separate diphoton pairs with signal-like kinematics, high photon ID scores and good mass resolution from background
Input variables:
Validation of Diphoton MVA is done on Z→ee events, with the electrons taken as photons
Higher BDT score gives better mass-resolution diphoton events

12. Event categorization, Signal and Background Modelling

13. Event classification
Events are split into different categories corresponding loosely to Higgs production modes
VBF- and ttH-specific categories with additional selection criteria to identify topologies corresponding to respective production modes
ttH: leptonic and hadronic modes
VBF: Split into 2 sub-categories based on sensitivity
VH was not integrated into workflow could not happen in time. To be added in the future.
4 Inclusive categories (“Untagged 0-3”), mostly consisting of ggH events : corresponding to different S/B and invariant mass resolution

14. Tagged events
ttH: leptonic tag
ttH: hadronic tag
VBF tag

15. Signal and background models
Fully parametric signal model from MC simulation
Background model data driven
physical nuisances allowed to float
corrections and data/MC efficiency scale factors applied
Background functional form treated as
discrete nuisance parameter
For each category, use different functional forms (sums of exponentials, sums of power law terms, Laurent series and polynomials)

16. Mass spectra : tagged events

17. Mass spectra : untagged events

18. Mass spectra : all categories
Unweighted
Weighted

19. Measurement results : significance, signal strength and couplings

20. Results : significance
Significance at GeV : 5.6σ observed (6.2σ expected)
Maximum observed significance is 6.1σ at GeV

21. Results : signal strength (I)
Per category
Per process
Best-fit signal strength
Signal strengths measurements compatible with SM

22. Results : signal strength (II)
Signal strength measured is measured in bosonic and fermionic components

23. Results : couplings
Measurement of coupling modifiers to vector bosons and fermions (kV, kf) and to photons and gluons (k,kg)

24. Fiducial XS measurement

25. Fiducial XS Measurement : Strategy
Fiducial region is defined using generator-level properties of particles: pTGEN/mGEN, ηGEN, IsolationGEN
Follows a similar strategy to the main analysis
Different event categorization: 3 mass resolution categories

26. Fiducial XS Measurement : Result
Fiducially cross section measured profiling mH:
Theoretical prediction for mH= GeV
Good data/theory agreement

27. Summary
CMS H→ results using 12.9/fb of 13 TeV collision data collected in 2016 have been presented
Observation (6.1σ peak significance) of the Higgs boson in the diphoton channel
Best fit signal strength is
Bosonic and fermionic components of signal strength are observed
Fiducial cross-section is measured to be
Consistency with SM

28. Thanks for your attention!

29. BACKUP

30. Analysis flowchart

31. Diphoton BDT
Inputs
Systematic uncertainties

32. ttH selection details Leptonic tag → Hadronic tag ↓
Diphoton MVA cut set to 0.5

33. Yields per event class : 12.9 fb-1

34. Systematic uncertainties
Theory uncertainties (PDFs, αS, QCD scale, underlying event and parton shower, H→ɣɣ branching fraction)
ggH contamination in VBF and ttH tagged categories
Trigger efficiency, integrated luminosity, vertex efficiency, preselection
Non-unformity of light collection, non-linearity, detector simulation, modeling of the material budget, shower shape corrections
Photon energy scale and resolution
BDT ID and per-photon energy resolution
Jet energy scale and smearings
b-tagging efficiency, gluon-splitting fraction, parton shower, ID efficiency for e and μ

35. Fiducial cross section : categorization
Categorization: 3 mass resolution categories
Categorization is validated on Z→ee with electrons reconstructed as photons

36. Mass : Run 1 combination

37. IHEP activities in H→ with 13TeV data
Small team in H→ working group : G. Chen, J. Tao, S. Zhang
One of the editors of the CMS internal docs for the measurements with both and TeV data (AN /PAS HIG , AN /PAS HIG )
Some contributions list
Run2 analysis framework (flashgg) development: photon validations codes and dimu EDM object construction
Diphoton HLT development : two developed HLT paths for low mass diphoton used in 2015 H→ to obtain good mass shape
Photon identification: Run1 proposed inputs; validations of the inputs and outputs with Z→  FSR photons
Selection efficiency: Photon selection efficiency and data/MC scale factors
Validations of shower shape corrections (R9, S4, etaWidth ) obtained from from Z→ee, with FSR photons from Z→ 
“一砖一瓦”