Calorimetry @ LHC (ATLAS & CMS)
Outline Introduction CMS EM calorimeter ATLAS EM calorimeter The problem of calibration CMS Hadronic calorimeter ATLAS Hadronic calorimeter
Large Hadron Collider: Higgs hunt Natural width (GeV) 0.001 0.004 1.4 30 250 0 50 100 200 400 800 Higgs Mass (GeV) LEP L3 H γγ LHC H ZZ * 4 leptons H ZZ 4 leptons H WW or ZZjj μ, e, γ LEP observed an excess of events around 115 GeV Only precision in γ detection will tell a peak (H γγ signal) from a huge background
Why precision matter so much? Response to monochromatic source of energy E H γγ bad resolution H γγ good resolution Perfect good bad background Calorimeter signal σ(calo) defines the energy resolution for energy E. m γγ Signal = constant integrated B σ γγ S/ B 1/ σγγ butσ γγ = f(σcalo)
ATLAS CALORIMETERS Hermetic system Electromagnetic Liquid Argon Calorimeters Tile Calorimeters η=1.475 η=1.8 η=3.2 Hadronic Liquid Argon EndCap Calorimeters Forward Liquid Argon Calorimeters
Em Barrel : EB Em Endcap : EE Had Barrel: HB Had Edcaps: HE Had Forward: HF Had Outer: HO CMS CALORIMETERS Hermetic system HB HO EB HF HE EE
ATLAS & CMS EM calorimetry Compact Excellent energy resolution Fast High granularity Radiation resistance E range MIP TeV Homogeneous calorimeter made of 75000 PbW0 4 scintillating crystals + PS FW Good energy resolution Fast High granularity Longitudinally segmented Radiation resistance E range MIP TeV Sampling LAr-Pb, 3 Longitudinal layers + PS ATLAS and CMS makes different choices: sampling calorimeter allow to have redundant mesurement of γ angle homogenous calorimeter with very low stochastic term aims to excellent energy resolution, the mesure of γ angle relies on vertex reconstruction from tracking.
m γγ = 2 E 1 E 2 (1 - cosθ γγ) 2 2 1/ 2 2 σm 1 σ 1 σ2 σθ σ( E) m = 2 E 1 H γγ: ECAL benchmark Γ H (m H 100 GeV) ~ 2 100 MeV Γ H /m H 10-3 + E 2 + Precision given by experimental resolution tgθ / 2 + E = a E b E c Homogeneus calo a can be ~ 2%, to match it for E γ ~ 50 GeV: Sampling calo a can be ~ 10%, to match it for E γ ~ 50 GeV: c ~ 0.5% b ~ 200 MeV CMS c ~ 0.7% b ~ 300 MeV ATLAS and an angular resolution σ θ ~ 50 mrad/ E and an angular resolution σ θ ~ 50 mrad/ E
ECAL @ CMS 75000 PWO crystals APD read out (gain 50) Eγ range 1 GeV 1 TeV Main technological challenges faced by ECAL-CMS: PWO: PbWO 4 about 10 m 3, 80 ton A change of scale! L3 BGO was 1 m 3 Development a suitable radiation hard crystal (PWO new scintillator) Light read-out in strong magnetic field (Avalanche Photo Diode new PD) Development of radiation resistant devices Production, test and assembly of 75000 crystals
Aiming at precision Precision has a price a long list to take care: Longitudinal and lateral shower containment Light production and collection Light collection uniformity Nuclear counter effect (leakage of particles in PD) Photo Detector gain (if any) stability Channel to channel intercalibration Electronic noise Dead material (energy loss and γ conversions) Temperature stability and uniformity Radiation damage Pileup
The choice of the crystal effectiveness & emission spectrum NO DETECTOR WITHOUT SUITABLE PHOTO_DETECTOR! Light Yield (LY) response time PD spectral sensitivity afterglow hygroscopic radiation length Hardness (Moh) Molière radius T t Radiaton damage Z eff ρ n
CMS developed a new crystal Lead Tungstate Crystals (PWO) for CMS 199 5 199 8 Very low light output Hard light extraction Very effective in high energy γ containment T dependent: -2%/ C 23 cm to contain em showers!
PWO: a scintillating crystal band gap Conduction band E g valence band E dep e-h E s = β E g β>1 N eh = E dep / βe g N γ = SQN eh Radiative efficiency of Efficiency of transfer luminescent centres to luminescent centres η γ = N γ /E dep =SQN eh /E dep = SQ/ βe g S, Q 1, βe g as small as possible medium transparent to λ emiss PbWO 4 : λ excit =300nm ; λ emiss =500nm intensity (a.u.) Stokes shift PWO 200 300 400 500 600 700 wavelength (nm)
PWO as grown Φ = 32 mm Ready for ECAL About 50000 crystals produced until now
Photon detectors for PWO Not sensitive to 4T magnetic field High quantum efficiency for λ 400 500 nm Internal amplification (low PWO LY) Fast and good for high rate (40MHz) Radiation hard Not (too much) sensitive to charged particles Photomultipliers affected by magnetic field large volume 200μm PIN photodiodes no internal amplification too sensitive to charged particles (Nuclear Counter Effect)
Avalanche Photo Diodes Barrel: Avalanche Photodiodes (APD, Hamamatsu) Characteristics optimized with an extensive R&D Program insensitive to B-field as PIN diodes Internal gain (M=50 used) good match to Lead Tungstate scintillation spectrum (Q.E. ~ 80%) dm/dv = 3%/V and dm/dt = -2.3%/ o C : T and V stabilization needed bulk current increase & recovery with irradiation measured over 1 year: expect doubling of initial noise after 10 years running, OK Capacitance 75 pf Excess noise factor F=2.2 ( fluctuations in multiplication) Effective d eff 6 μm ( small response to ionizing radiation) 2 APDs per crystal: 50 mm 2 active area
Electronic system Designed to preserve signal information VFE x 5 FE Front End card (FE) Trigger Sums MB LVR Data rigger Tower (TT) Very Front End card (VFE) HV 2 x12 1 Logic x6 12 bit ADC MGPA 0 x1 APD/VPT 12 bits 2 bits VFE architecture for single channel Trigger primitives computed on the detector Command&control viatokenring Modularity: Trigger Tower (25 channels in Barrel) - 1 Low Voltage Regulation Board (LVR) - 5 VFE Boards (5 channels each) - 1 FE Board - 1 Fibre sending trig primitives (every bunch Xing) - 1 Fibre sending data (on Level1 accept)
Energy resolution: a, b, c In scintillating crystals the only intrinsic source of fluctuations is photostatistics: Light Yield of the crystal is one of the factors but not the only one σ E = 1 N pe = 1 E(GeV) N pe GeV Npe/GeV= (γ/gev) (light collection eff.) (geometrical PD eff.) (photocathode eff.) a = (photostatistics) (lateral containment) (e multiplication in PD) Electronic noise (1/E): b = (pd capacitance) (dark current) (physics pileup) 1/ t shaping t shaping c = (leakage) (intercalibration) (system instability) (nonuniformity of xl) To have c 0.5 % all contributions must stay below 0.3 %
CMS ECAL: the performance 1 Super Module 1700 xl on test beam in 2004 Noise distribution Average resolution : 30 MeV 45 MeV
Energy resolution: constant term Intercalibration requires several steps before, during and after data taking test beam precalibration continuous monitor during data taking absolute calibrations by physics reactions during the experiment lifetime THIS IS THE KEY ISSUE TO MAINTAIN PHYSICS PERFORMANCE
Things may change unexpectedly 1.02 AEGING GOING ON L3 BGO 1 1990 1991 RB26 (Hb 1) RB24 (Hb 2) 0.98 1992 System able to track the BGO response decrease (few %/year) with light injection Porting of previous year calibration:1.3% Spread after Xe+Bhabha corrections: 0.8% from calibration in 1991 Electron energy/beam energy 0.96 0.94 0.92 0.9 Barrel 1993 1994 1991 0.88 (ageing of some optical component) 0 200 400 600 800 1000 1200 1400 1600 1800 Time (days) In 1999 0.5% from calibration after refinements of methods
L3 BGO ECAL: calibration Results on 45 GeV Bhabha electrons (after continuous refinement of methods) 1991 1.25% (0.8 from calibration) 1999 1.06% (0.5 from calibration)
LY irr /LY 0 (%) 105 100 95 CMS PWO γ induced radiation damage Front irrad., 1.5Gy, 0.15Gy/h 90 0 0.5 1 1.5 2 2.5 E. Auffray, EP_CM PWO_batch06lowdoselab27.qpc Dose (Gy) 18/01/2000 Low dose rate irradiation of some BTCP crystals of Batch06 in lab27 γ PWO4510 (%LY) PWO4579 (%LY) PWO4585 (%LY) PWO4590 (%LY) PWO4622 (%LY) PWO4623 (%LY) PWO4533 (%LY) PWO4481 (%LY) PWO4473 (%LY) Dose (Gy) We know PWO response will change with irradiation! Simulation of crystal transparency evolution at LHC (L =2x10 33 cm -2 s -1 ) - based on test beam irradiation results The Problem: Colour centres form in PWO under irrad n Transparency loss depends on dose rate Equilibrium is reached after a low dose Partial recovery occurs in a few hours
CMS ECAL monitoring system The Solution: Damage and recovery during LHC cycles tracked with a laser monitoring system 2 wavelengths are used: 440 nm and 796 nm Light is injected into each crystal Normalisation given by PN diodes (0.1%) APD CRYSTAL (1700/SM) Monitoring of evolution by light injection system PN (200 Channels) LEVEL-1 LEVEL-2 FANOUT FANOUT SWITCH (select SM/2) LASER
ECAL monitoring system Relation expected between S (beam signal) and R(laser signal) S cor = S R R 0 α Experimental determination of α: lns = α ln(r/r 0 ) + K where K= ln(s 0 ) beam laser NB: α is the same for all crystals!
ECAL monitoring system Fixing mean α Beam reconstructed response S cor is computed from experimental data S cor = S R R 0 α 0.5 Gy/h 0.16 % 0.2 % Before corrections After corrections Irradiation Recovery Laser monitor provides a good compensation over short/long period HAVE TO BE PROVEN DURING DATA TAKING ON THE WHOLE CALORIMETER
ECAL @ ATLAS φ Sampling: accordion lead structure filled with LAr 1 module covers η: 0 to 1.4, φ: 0.4 Longitudinal dimension: 25 X 0 = 47 cm (CMS 22 cm) 3 longitudinal layers 4 X 0 π 0 rejections separation of 2 photons very fine grain in η 16 X 0 for shower core 2X 0 evaluation of late started showers Total channels 170000 Particles from collisions
ATLAS: the choice of LAr High number of electron-ion pair produced by ionization No amplification neeeded of signal, low fluctuations Liquid Very uniform response (purification) Stability with time Main fluctuations are due to sampling fluctuations Intrinsically radiation hard cheap slow time response 400 ns boling temperature 87 K criogeny needed Temperature sensitivity 2% signal drop for ΔT=1 C
ATLAS EM LAr HT I phys Signal is given from collection of released electrons Gerbe EM e - Drift velocity depends on electron mobility and applied field. In ATLAS : Lar gap 2 mm, ΔV = 2kV e - γ ions e - Electrode 400 ns 16 LHC BC e + Plomb Argon liquide Signal E ~ 1kV/mm After shaping Pulse is shaped and sampled each 25 ns, has 0 time integral mean value of pileup is cancelled (no baseline shift).
LAr electronics calibration The ionization signal is sampled every 25 ns by a 12 bits ADC in 3 gains. 5 samples are recorded at at ATLAS. The shaper output of the ionisation and calibration signal is different! Time Amplitude ( Energy) sampled at 40 MHz and digitised Pedestal subtracted Injected signal shape NEED Different Injection point CORRECTIONS The equalization of the electronic readout. Requires to know the shaping function of each cell at few percent level equalization with an electronic control signal
<>= 2.211 mm σ =10 μm The challenge of LAr Absorber thickness Mechanical non uniformities: modifies electric field and detector response. Take care during construction, try to reproduce effects and apply corrections. 1% Pb variation 0.6% drop in response Measured dispersion σ = 9 μm (calo) translates to < 2 effect on constant term Response to 120 GeV e-showers slant angle : 1º/~100º is sensitive φ-modulations in the EMEC EM calorimeter : Pb absorbers Peculiar accordion shape sagging Calorimeter response is affected ~ 3 % an as built detector : HV, sagging, misalignment
ATLAS EM uniformity Module P13 245.6 GeV Module P15 245.7 GeV 0,44% 0,44% Uniformity 0,7-0,9% Scan modules with monochromatic electrons 0,7-0,9% Resolution Module P13 P15 Global constant term 0.62% 0.56% P13/P15 ~ 0.05% Ratio of absolute response
ATLAS EM: the performance TB2002 Local result? The constant term in the resolution is dominated by: the equalization of the electronic readout. the non uniformity in the electric field and in the sampling fraction introduced by the accordion structure.
The calibration From single channel electrical signal to E e,γ The case of CMS because is more easy to understand TBD E i Cluster absolute energy scale amplitudes inter-calibration constants algorithmic corrections (particle type, momentum, position & clustering algo) Account for energy losses due to containment variations
The tough point: material in Trackers CMS + THE SOLENOID Tough for both experiments ATLAS 1 X/X 0 0.5 Tracker material : electrons loose energy via bremsstrahlung photons convert 4T (2T) solenoidal B field : Electrons bend radiated energy spread in φ γ e -3 η -1 0
Calibration: effect of material EFFECT IN CMS intrinsic ECAL resolution: 0.7% 50% e not negligible brem definition of algorithm and selection efficiency for e with no brem e track reconstruction e reconstruction quality f(η,φ) effect of initial calibration on reconstruction and selection (SuperCluster from dynamic clustering algorithms) super-cluster basic cluster The size of the tail is eta depending!
CMS calibration: @ start up 50000 xls 12% spread ~ 4.2% LY measured in Regional centers (CERN, INFN-ENEA Rome) with 1 MeV γ ( 60 Co) use cosmics Comparison with TestBeam : 4.2% inter-calibration precision. Compare cosmics intercalibration (raw) with test beam results: agreement at 3% (will improve) START WITH MASS INTERCALIBRATION KNOWN @ 3-5 % + FEW REFERENCE SM @ 0.5% (TEST BEAM HIGH E electrons)
CMS relative and absolute calibration USE PHYSICS EVENTS DURING DATA TAKING (W eν, Ζ e + e - ) Method: Z mass constraint Method: E / P <width minimization> σ Z M σ cal 2 Z N electrons 2.0 fb -1 ECAL 5x5 E = Σ c i Ε i TRACKER electron momentum But also: π 0 γγ, η γγ, Z μ + μ - γ, Jets/min-bias (φ-uniformity of deposited energy in crystals at constant η)
And the Higgs? If light, it will take a while Relative Higgs mass resolution versus mis-calibration. Higgs Boson Mass Resolution H γγ Barrel On paper resolution on γ γinvariant mass: CMS 0.7 GeV ATLAS 1.2 GeV
HCAL @ ATLAS Hadronic Tiles Barrel (Liq Arg EM calorimeter cryostat) (Forward calorimeters cryostats) Hadronic Tiles Extended barrel z (or η) Tiles perpendicular to beam axis Wavelength shifting fibers carry light to PMTs Covers η <1.7 Hadronic Calorimeter: Iron/Plastic scintillator sampling calorimeter
ATLAS HCAL Linearity Energy resolution σ = E 41.9% E 1.8 + 1.8% E
ATLAS HCAL UNIFORMITY OF RESPONSE 180 GeV π beam RMS/Mean = 1.8%
HCAL @ CMS HCAL Outer HO
CMS HCAL ~ 5% of a 300 GeV π energy is leaked outside the HB (inside coil) HB inside the coil not enough thick for shower containment: scintillator layers just after the coil (HO) improves π resolution by ~10% at 300 GeV & linearity
CMS HCAL π interacting in HCAL σ = E 101% E 4% Effect of different e/h + no longitudinal sampling in EM σ = E 122% E 5%
HCAL: compare parameters
HCAL The choices made for the hadronic central section by ATLAS and CMS are similar: sampling calorimeters with scintillator as active material. In both cases the dominant factor on resolution and linearity is the e/h 1 ATLAS & CMS: e/h had 1.4 ATLAS higher segmentation and containment gives better total resolution
Missing E T : expected performances SHOULD BE 0 IN QCD EVENTS
Physics objects We are not going to measure single hadrons Contribution from Physics: Parton shower & fragmentation Underlying events Initial State Radiation & Final State Radiation Pileup form minimum bias events Detector: Resolution Granularity Clustering: Out of cone energy losses Use physics events to understand jet energy reconstruction: γ / Z ( ll) + jet, W jet jet,...
Conclusions Many important arguments have been left out, I made a choice somewhat complementary to last year lectures (C. Roda). Many people spent a lot of time and effort to realize these projects, now physics is near and I hope you will profit and help ATLAS and CMS to optimize detectors and algorithms, to make an harvest of discoveries and to interpret them. THANK YOU FOR YOUR ATTENTION
Few references R. Wigmans, Calorimetry, Energy Measurements in Particle Physics Priscilla B.Cushman, Electromagnetic and Hadronic Calorimeters U.Amaldi, Fluctuations in Calorimetry measurements 1981 Phys.Scr.23 409 C.W.Fabjan and F.Gianotti, Calorimetry for particle physics, Reviews of Modern Physics, Vol.75, October 2003 ATLAS & CMS TDRs