Role of the CMS electromagnetic calorimeter in the measurement of the Higgs boson
properties and search for new physics
Federico Ferri CEA/Saclay Irfu/SPP on behalf of the CMS Collaboration
ICHEP 2014, Valencia, July 2-9 2014
Introduction . . . . what you already know
Understanding the detector . . . . what you already heard
Performance and its evolution . . . . what you would like to see
Some detail on systematics effects . . . . what you wonder about
Conclusions . . . . 12 min over, more at coffees
Preamble
Homogeneous, hermetic, high granularity PbWO
4crystal calorimeter
density of 8.3 g/cm
3, radiation length 0.89 cm, Moli` ere radius 2.2 cm, ≈ 80% of scintillating light in ≈ 25 ns, refractive index 2.2, light yield spread among crystals ≈ 10%
Barrel: 61200 crystals in 36 super-modules, |η| < 1.48, Avalanche Photo-Diode (APD) readout
Endcaps: 14648 crystals in 4-Dees, 1.48 < |η| < 3.0, Vacuum Photo-Triode (VPT) readout
Preshower (endcaps only): 3X
0of Pb/Si strips, 1.65 < |η| < 2.6
Solenoidal magnetic field: 3.8 T ECAL fully contained in the coil
CMS tracker coverage: |η| < 2.5
Physics with the ECAL
Design emphasis: energy resolution and granularity [1]
Byproduct: time resolution
Successful examples so far:
0 0.5 1 1.5 2 2.5 3 3.5
4 19.7 fb-1 (8 TeV) + 5.1 fb-1 (7 TeV) CMS
103
×
(GeV) γ mγ
110 115 120 125 130 135 140 145 150
-100 0 100
200 B component subtracted
S/(S+B) weighted events / GeV
S/(S+B) weighted sum Data S+B fits (weighted sum) B component
σ
±1 σ
±2
H → γγ
energy resolution, position reconstruction, granularity for photon identification
H → ZZ → 2e 2µ(4e ) good reconstruction of electrons with very low E
TLong-lived neutralino physics reach significantly extended by using the ECAL timing information
[email protected] ICHEP 2014, Valencia, July 2-9 2014 3
ECAL performance at Test Beams
Perfect calibration, no magnetic field, no material upstream, negligible irradiation
Energy resolution
central impact, 3×3 barrel crystals [2] :
σ(E)
E = 2.8%
√ E ⊕ 0.128
E(GeV ) ⊕ 0.3%
constant term to be kept 1%
stochastic term also affected by the material upstream
Time resolution: constant term ≈ 20 ps
from time difference of crystals in the same eγ shower
Challenges of in situ operations
Light yield variations:
scintillation light → temperature dependence: ∆S/S ∼ −2%/
◦C @ 18
◦C crystal transparency → radiation dose-rate dependence
Photo-detector response:
APD →
gain temperature dependence: ∆G/G ∼ −2%/
◦C gain High-Voltage dependence: ∆G/G ∼ 3%/V direct ionization effects, a.k.a. “spikes”
VPT → response dependence on the incremental charge at the cathode Tracker material in front of ECAL:
photon conversions
bremsstrahlung losses for electrons 3.8 T solenoidal magnetic field:
spread of the e, γ energy along ϕ, at ≈ constant η
→ Excellent environmental stability (×2 to ×3 better than required) [3]
→ Dedicated monitoring system and calibration techniques [4, 5]
→ Specific energy reconstruction algorithms and corrections
[email protected] ICHEP 2014, Valencia, July 2-9 2014 5
Ingredients for precision physics
Electrons and photons deposit energy over several crystals (70% in one 97% in a 3×3 array), spread in ϕ, collected by “clustering” algorithms
E e,γ = G F e,γ
X
i
c i s i (t)A i
A i : single channel amplitude , optimal filter in the time domain
s i (t) : single-channel time-dependent response corrections , via a dedicated laser monitoring system
c i : inter-calibration of the single channel response, using physics: ϕ- and time-invariance of the energy flow in minimum-bias events, π
0, η → γγ and Z → ee invariant mass peak, electron E/p
F e,γ : particle energy correction (geometry, clustering, . . . )
G : global scale calibration, with Z → ee events
Resolution, efficiency and particle ID: Z → ee
Monitoring and calibration signals
Laser monitoring measurements Validation of the corrections with E/p
Inter-calibration precision
[F. De Guio’s talk]
EB RMS: 0.09%
EE RMS: 0.3%
[email protected] ICHEP 2014, Valencia, July 2-9 2014 7
Performance: time resolution
Better than O(1 ns) stability required for precise energy determination → regular calibrations
Fast scintillation response (≈ 80% of light within 25 ns) , shaping time (≈ 40 ns) , and sampling rate (40 MHz) allows for excellent time-resolution
From the time difference between the highest energy crystal of each of the two electrons from a Z → ee
Noise term consistent with Test-Beam
Constant term of ≈ 150 ps, much better than design, uniform and stable in time
residual differences with Test-Beam
qualifications ascribed to the clock
distribution system
Performance: energy resolution
With electrons from Z
↑ Fit to Z → ee of a Breit-Wigner convolved with a Gaussian function [4]
→ Simulation tuned to match performance observed in situ with Z → ee events
Events / 0.5 GeV
0 5 10 15 20 104
× 19.7 fb-1 (8 TeV)
CMS
Barrel-Barrel Data
(MC) e- e+
→ Z
(GeV) mee
75 80 85 90 95 100 105
Data/MC 0.8 1 1.2
Events / 0.5 GeV
0 20 40 60 80 103
× 19.7 fb-1 (8 TeV)
CMS
Not Barrel-Barrel Data
(MC) e- e+
→ Z
(GeV) mee
75 80 85 90 95 100 105
Data/MC 0.8 1 1.2
[email protected] ICHEP 2014, Valencia, July 2-9 2014 9
Accurate simulation
Noise model:
realistic noise with sample-correlations and channel-to-channel variations
increase of the APD dark current (expected)
transparency variations for realistic light-yield (and corresponding photo-statistics)
Material description:
including in-homogeneities in ϕ of services in front of the endcaps
for systematic uncertainties, being implemented in current simulation
η|
|
0 0.20.40.6 0.8 1 1.21.4 1.61.8 2
MC 0/Xdata 0X
0.9 1 1.1 1.2 1.3 1.4 1.5
Hadron track momentum loss Electron track Weighted average
= 8 TeV s CMS preliminary
in )/p out -p in (p
Light propagation effects in the crystals (only relevant for upgrade studies)
Varying conditions used for a “run-dependent” simulation [N. Marinelli’s talk]
Performance evolution
The energy resolution measured in data with Z → ee is used to model the expected H → γγ signal in the simulation
Steady progress and excellent results
Arabella Martelli TIPP14 02/06/14
Evolution of ECAL understanding...
‣ Expected H->γγ signal modelled with energy resolution measured from data (Z->ee)
‣ PROMPT reconstruction within 48h from data taking
‣ RECONSTRUCTION with improved conditions
10 15
2) (GeV/c
! m!
100 110 120 130
)2Events / ( 0.35 GeV/c
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 Simulation
Parametric Model
= 2.45 GeV/c2
"eff
Simulation CMS preliminary
Combined All Categories
FWHM = 4.35 GeV/c2
Figure 8: Simulated signal (mH=120 GeV/c2) for all 8 event classes combined. The line shows the weighted sum of fits to the signal models in the individual classes.
8 Limit setting
The confidence level (CL) for exclusion or discovery is evaluated using the diphoton invariant mass distribution as the observable for each of the event classes defined in Section 6. The results in the 8 classes are combined in the CL calculation to obtain the final result.
Two statistical approaches are considered in evaluating limits: the modified frequentist ap- proach (CLS) using the profile likelihood as a test statistic [18], and a Bayesian approach with a flat prior for the signal strength. These two methods are generally expected to give similar results and so provide a valuable cross check of the statistical procedures.
Both a binned and an unbinned evaluation of the likelihood are considered. While most of the analysis and determination of systematic uncertainties are common for these two approaches, there are differences at the final stages which make a comparison useful. The signal is model taken from the MC after applying the corrections determined from data/Monte Carlo compar- isons ofZ eeandZ µµmentioned above, and the reweighting of thepHTspectrum. In the unbinned evaluation the signal model is parametric, based on analytic functions fitted to the Monte Carlo, whereas the binned evaluation uses templates made with Monte Carlo events.
The comparison of results thus verifies that the parametric model describes the Monte Carlo well. For the background, Monte Carlo is not used and the background is evaluated from a fit to the data.
Given the narrowness of the Higgs mass peak which has a resolution approaching 1 GeV/c2in the classes with best resolution, the search must be carried out in fine steps. At present steps of 500 MeV/c2are used.
All known sources of relevant systematic uncertainties have been described in the previous sec- tions. Table 6 lists systematic uncertainties on the signal applicable to the individual photons,
2) (GeV/c
! m!
100 110 120 130
)2Events / ( 0.35 GeV/c
0 1 2 3 4 5 6 7 Simulation
Parametric Model
= 1.76 GeV/c2
"eff
Simulation CMS preliminary
All Categories Combined
FWHM = 3.18 GeV/c2
(GeV)
! m! 105 110 115 120 125 130 135 140
Events / ( 0.5 GeV )
0 10 20 30 40
50 Simulation
Parametric model
= 1.84 GeV
"eff
FWHM = 3.05 GeV
Simulation CMS Preliminary
Combined (GeV)
! m!
105 110 115 120 125 130 135 140
Events / ( 0.5 GeV )
0 5 10 15 20 25 30 35
40 Simulation
Parametric model
= 2.02 GeV
"eff
FWHM = 3.86 GeV
Simulation CMS Preliminary
All Categories Combined
7TeV 8TeV
ICHEP 2012 2014
FWHM 2.35
FWHM 2.35
FWHM 2.35 FWHM
2.35
= 1.54%
= 1.12% = 1.04%
= 1.31%
LP 2011
MORIOND 2013
[email protected] ICHEP 2014, Valencia, July 2-9 2014 11
ECAL-related systematic uncertainties on m H
From H → γγ:
m H = 124.70 ± 0.31(stat) ± 0.15(syst) GeV
Electron/photon differencies in the simulation . . . . 0.10 GeV
material distribution. . . . 0.07 GeV
longitudinal light-yield non-uniformity . . . . 0.02 GeV
Geant4 . . . . 0.06 GeV
? uncertainty on the single contribution: ≈ 10 MeV
N.B.: the detector response to electrons and photons shows differences at the level of 0.5%. What matters is the difference of these differences between data and simulation.
Residual non-linearity in scale . . . . 0.10 GeV
Photon energy scale corrections . . . . 0.05 GeV
Z line shape . . . . 0.01 GeV
Checked and negligible contribution: gain switch of the electronics
More detail: residual non-linearity in scale
/2 (GeV) or H
TE
T20 40 60 80 100
Relative response
0.998 1 1.002 1.004CMS Unpublished EB-EB high R9
/2 (GeV) or H
TE
T20 40 60 80 100
(8 TeV) 19.7 fb
-1EB-EB low R9
/2 (GeV) or H
TE
T20 40 60 80 100
Relative response
0.99 1 1.01
E/p data/MC EE-EE high R9
/2 (GeV) or H
TE
T20 40 60 80 100
data/MC m
eeEE-EE low R9
Residual non-linearity of the energy response in data relative to simulation, relevant in the extrapolation from the energy scale measured at the Z peak (≈ 90 GeV) to the Higgs boson mass (≈ 125 GeV)
1. electron E/p vs. E
Twith electrons from Z and W decays 2. di-electron invariant mass vs. H
T= E
T1+ E
T2in Z → ee events
0.08% effect on the Higgs boson mass
[email protected] ICHEP 2014, Valencia, July 2-9 2014 13
More detail: longitudinal non-uniformity (NUF)
R&D achievements: adequate uniformity of longitudinal light yield
one face of each barrel crystal depolished
X (mm)
0 100 200
Energy collection weight
0.96 0.98 1 1.02 1.04
dE/dX (GeV/mm)
0 0.2 0.4 0.6 0.8 dE(e)/dX (E=60 GeV)
)/dX (E=60 GeV) γ dE(
Default profile Modified profile CMS
Unpublished
Simulation: rear non-uniformity of 0.15%, front part assumed uniform
Ionizing radiation found to induce additional NUF of 30% of its initial value (worst case scenario) at the end of Run1 [6]
→ simulation modified to account for these effects
at most 0.06% effect on the energy scale , anti-correlated between
converted and un-converted photons → 0.015% effect on the mass
Conclusions
The CMS ECAL has been deeply understood and performance optimized, thanks to an endless effort towards improvement and a careful scrutiny of all the details
Essential role in the Higgs boson prompt discovery and precise mass measurement http://arxiv.org/pdf/1407.0558.pdf
The concepts that have driven the design of the detector more than 20 years ago have proven to be successful
and the meticulous characterization of all the crystals in laboratory has paid back with reduced systematic uncertainties
Looking forward to start Run2 to exploit the excellent discovery potential for physics beyond the Standard Model
with new challenging conditions: increase of collision energy and pile-up, different machine parameters. . .
[email protected] ICHEP 2014, Valencia, July 2-9 2014 15
Bibliography
[1] CMS-ECAL Technical Design Report, CERN/LHCC 97-33 (1997)
[2] P.Adzic et al. (CMS ECAL), Energy resolution of the barrel of the CMS Electromagnetic Calorimeter, J.Inst. 2 P04004 (2007)
[3] CMS Collaboration, Performance and operation of the CMS electromagnetic calorimeter, J.Inst. 5 T03010 (2010)
[4] CMS Collaboration, Energy calibration and resolution of the CMS electromagnetic calorimeter in pp collisions at √
s = 7 TeV, JINST 8 (2013) P09009
[5] M.Anfreville et al. (CMS ECAL), Laser monitoring system for the CMS lead tungstate crystal calorimeter, Nucl.Instrum.Meth. A594 (2008) 292-320
[6] F. Nessi-Tedaldi, Response Evolution of the CMS ECAL and RD Studies for
Electromagnetic Calorimetry at the HL-LHC, IEEE Trans.Nucl.Sci. 60 (2013)
pp.2831-2893
