1. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 1 David Futyan Imperial College Calibration of the CMS ECAL Using Vector Bosons Calibration of the CMS ECAL Using Vector Bosons

2. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 2 Overview Introduction: The CMS ECAL Calibration requirements and strategy at LHC startup Precalibration: Laboratory measurements Testbeam Cosmic Rays In-situ calibration with physics events: Phi independence Vector bosons: Single electrons: W  e Double electrons: Z  ee Single photons: Z  Low mass resonances:  0 

3. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 3 The CMS Detector ECAL

4. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 4 The CMS ECAL 75848 Lead Tungstate crystals. Barrel geometry: Front face dimensions: 0.0174  0.0174 in  (22  22mm = Molliere radius) Crystal depth: 25.8X 0 (230mm) Crystal axes tilted by 3 o w.r.t. line from nominal vertex 3X 0 preshower in front of most of endcap    1.29m

5. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 5 The CMS ECAL Barrel “supermodule” (1700 crystals) composed of 4 modules. Each Half barrel contains 18 supermodules. Readout: Barrel: avalanche photo-diodes (APDs) Endcap: vacuum phototriodes (VPTs) Photomultipliers with single gain stage Able to operate in 4T magnetic field and high neutron flux

6. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 6 ECAL Energy Resolution Physics reach of the ECAL, in particular the H  discovery potential, depends on its excellent energy resolution. Achievement of deign performance requires high precision calibration. Intrinsic ECAL energy resolution of the CMS ECAL: Constant term dominated by intercalibration precision (most of the energy of an electron or photon goes into a single crystal)

7. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 7 Calibration Requirements Challenge is the relative channel-to-channel intercalibration of the ~80K crystals. Intercalibration precision feeds directly into the constant term Global absolute energy scale can be obtained using a relatively small no. of Z  e + e - or Z  events. Main source of channel-to-channel response variation: Barrel: crystal scintillation light yield, RMS ≈ 8% Endcaps: VPT signal yield, RMS ≈ 25% Aim is to achieve an intercalibration precision of better than 0.5%.

8. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 8 Calibration Requirements Ultimate goal is to achieve the most accurate energy measurement for electrons and photons: G x c i x A i are calibrated RecHits G = global absolute scale c i = calibration coefficients A i = signal amplitudes in ADC counts F = cluster level energy corrections to correct for energy loss due to bremsstrahlung and containment variations dependent on type of particle, position, momentum, clustering algorithm…

9. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 9 ECAL Intercalibration Strategy at LHC Startup Preliminary estimates of intercalibration coefficients: Laboratory measurements of crystal light yield Test beam precalibration of some supermodules Commissioning of further supermodules with cosmic rays Target precision can only be achieved in-situ using physics events: Impose  -independence of energy deposited from minimum bias or jet triggers to rapidly intercalibrate to a precision of around 2%. Intercalibrate between regions using Z  e + e -. Finally intercalibrate to design goal of <0.5% using the momentum measured in the tracker for electrons from W  e (requires tracker alignment to be complete) Complimentary method, not relying on tracker: invariant mass reconstruction from    and  resonances.

10. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 10 Precalibration

11. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 11 Laboratory Measurements Excite crystal with a 60 Co source. Measure light yield (LY) with PM tube. Prediction of calibration coefficient c i :  Q = photo-detector quantum efficiency c ele = electronics chain calibration M = photo-detector gain Precision of lab measurement of c i can be determined by comparing to c i determined from testbeam:  4%

12. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 12 Testbeam Precalibration 8 supermodules (~1/4 or ECAL barrel) precalibrated in summer 2006 Remaining 28 barrel supermodules will be intercalibrated in-situ One of the four endcap “Dees” will be precalibrated in the summer 2007 testbeam Full scan of supermodule using high energy electron beam Crystal response depends on electron impact position. 4th order polynomial, separately in the 2 lateral coordinates used to correct for this dependence. Only electrons incident on central 7mm  7mm of a given crystal are used. c i defined as ratio of mean value of corrected response w.r.t. reference value Statistical uncertainty with 1000 events/crystal < 0.1% Intercalibration precision limited by response variations during time between testbeam measurement and LHC data taking. Repeat precalibration of one module to quantify reproducibility

13. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 13 Calibration Using Cosmic Ray Muons Calibrate using cosmic muons well aligned to crystal axes. Select events with nearly all energy in one crystal Supermodule exposed to cosmic muons for 41 hours in Nov 2004 Precision of 3% achievable in 1 week of data taking for barrel modules 1-3, and 3.5% for module 4. datasimulation Agreement with test beam calibration:  3%

14. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 14 In-Situ Intercalibration Using Physics Events 1: Phi-Independence

15. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 15 Phi Independence  -symmetry of event activity  Total E T deposited by a large number of events should be the same for all crystals at a given .  Can perform intercalibration by comparing  E T deposited in a crystal with the mean  E T for all crystals at the same . Aim: reduce the number of intercalibration constants from ~76000 (number of ECAL crystals) to 248 (number of fixed  rings). Rings can then be intercalibrated using Z  e + e - events.  y x Invariant quantity is  E T /A  for endcap rings 170 barrel rings 39 endcap ring pairs

16. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 16 Phi Independence: Event trigger and Selection Event trigger: Level-1 Jet trigger Alternatives considered: Random bunch-crossings (minimum-bias): No trigger bias from the event trigger, but sensitivity to noise due to low energies, and large extrapolation from calibration energies (few hundred MeV) to physics energies Electromagnetic triggers: Trigger bias a severe issue Assume 1kHz of Level-1 bandwidth allocated to single jet triggers Event selection consists only of an E T threshold: E T > 120 GeV, chosen such that the 1kHz bandwidth is approximately saturated at LHC startup luminosity.

17. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 17 Only crystals with E T in the following ranges contribute to the E T sums: 1 < E T < 6 GeV (barrel) 1 < E T < 4 GeV (endcap) ~10 crystal hits per event pass this selection. Write out highly compacted data stream to be processed offline: Only information stored for each event is energy and crystal ID for each selected hit Lower threshold excludes noise; upper threshold removes low statistics tail (improve stability of E T sum) Eliminate trigger bias by excluding crystals associated with the triggering (highest E T ) jet: Require that crystals are separated from the position of the triggering jet by:  R =  (  2 +  2 ) > 1rad Selection of Crystal Energy Deposits

18. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 18 Calculate  E T (  E T /A  for endcap) for each crystal. Obtain  E T (mean value of  E T for all crystals in pair of rings). For each crystal: Calculate fractional deviation of  E T from  E T : Since  E T is obtained from a truncated E T distribution,  is proportional rather than equal to the miscalibration. Constant of proportionality k determined empirically for each pair of rings (value is typically ~1.5). Estimate of miscalibration: Calibration coefficient: To test calibration procedure: Gaussian miscalibration applied with spread 4.5% Determine residual miscalibration after correction Determination of Calibration Coefficients 11 million events

19. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 19 Intercalibration Precision Assumes no knowledge of tracker material distribution Limit of the technique reached when inhomogeneity of tracker material and crystal geometry breaks  -symmetry of energy deposition.

20. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 20 Limit is approached with a few tens of millions of events - equivalent to around 10 hours of data taking assuming 1kHz of Level-1 bandwidth allocated to single jet triggers. With increasing knowledge of tracker material distribution, potential for rapid repeated calibration of the ECAL to high precision. Limit on Precision

21. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 21 2: Single Electrons from W Bosons

22. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 22 Calibration with W  e : Introduction Intercalibrate using the tracker momentum measurement for isolated electrons Benchmark technique to achieve target precision (0.5%) Requires tracker is fully operational and well aligned W  e : source of single electrons with a high HLT rate of ~10 Hz at 2x10 33 cm -2 s -1 Main difficulty: Bremssrahlung radiated in tracker material degrades electron energy and momentum measurements <10% radiated energy No cut on radiated energy E(5x5)/p(track)

23. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 23 ECAL Energy Measurement ECAL energy measured by summing 5x5 array of crystals around crystal with maximum signal. In the endcaps, energy deposited in the preshower is added Choice of 5x5 rather than clustering algorithms designed to recover bremsstrahlung in standard electron reconstruction: Cleanly separate intercalibration from complex algorithmic corrections required for bremsstrahlung recovery Energy in 5x5 gives best measurement of energy for unconverted photons, for which calibration accuracy is most important (H  )

24. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 24 Regional Calibration Amount of bremsstrahlung depends on amount of material varies significantly as a function of . Average value of E/p distribution therefore also varies with . Divide intercalibration task into 2 steps: 1) Intercalibrate within small  regions for which is rather constant Can be achieved rapidly for each region due to the reduced no. of constants to be determined 2) Intercalibrate between regions. Use very tight electron selection requiring minimal bremsstrahlung energy loss

25. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 25 Event Selection Event selection based on variables correlated with amount of bremsstrahlung emission. Electrons with little or no radiated bremsstrahlung have: More accurate reconstruction of energy and momentum Most of their energy deposited in 5x5 crystal array Barrel selection variables: E(5x5)/p(track), plus: E(3x3)/E(5x5) No. of hits in the track  2 /n.d.f. of the track

26. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 26 Endcap Event Selection Selection variables used in the endcaps: E(5x5)/p(track) E(3x3)/E(5x5) Ratio of track momentum at outermost and innermost points: p out /p in Fraction of energy radiated before radius = 80cm

27. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 27 Selection Cut Optimization Selection cuts chosen by scanning 4 dimensional phase space and identifying point (global minimum) giving best precision in determination of calibration constants. Global minimum found to be broad i.e. stable - insensitive to potential differences between simulation, used to derive the cuts, and real data used to perform calibraition Sensitivity of minimum on no. of events per crystal also found to be small Selection cuts optimized separately for the different  regions Selection efficiency varies with  Mean efficiency in barel: 30% Efficiency in endcap ranges from 10% to 30% E(5x5)/p(track) (Endcap) Selected events

28. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 28 Calibration Algorithms Each energy measurement contains contribution from 25 crystals. Two techniques have been studied to extract calibration constants for individual crystals: 1) L3 iterative algorithm: Used for in-situ calibration of the BGO crystals in the L3 experiment at LEP 2) Matrix inversion algorithm: Householder decomposition or  2 minimization Based on minimization of difference between E ECAL and p track Both techniques perform similarly, both in terms of precision and speed. L3 iterative algorithm

29. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 29 Intercalibration Precision To test calibration procedure: Gaussian miscalibration applied with spread 4%. Many (>50) MC experiments performed each with different randomly chosen miscalibration constants Intercalibration precision vs  : barrel endcaps

30. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 30 Intercalibration Precision Dependency on no. of electrons per crystal: 1.31<  <1.48 0.78<  <0.96 0<  <0.26

31. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 31 HLT Rates and Background Contamination Most of the background rate comes from b/c  e decays. Since these are real electrons, can still be useful for calibration Otherwise Can be strongly suppressed with negligible loss of signal efficiency using isolation cuts W  e background No isolation cut

32. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 32 3: Electrons from Z Bosons

33. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 33 Calibration with Z  ee: Introduction Z mass constraint in Z  ee events is a powerful tool for ECAL calibration Phi independence and W  e concerned only with intercalibration. Z  ee can also be used to determine the global energy scale. Independent of the tracker measurements: can be used from the beginning of the data taking. Several uses are envisaged: Intercalibration between rings previously intercalibrated using phi- independence. Determination of global energy scale. Tuning of algorithmic (clustering) energy scale corrections for reconstructed electrons and photons, currently determined using Monte Carlo. Measurement of electron trigger, reconstruction and identification efficiencies using “tag and probe” approach.

34. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 34 Extraction of Calibration Coefficients Measured invariant mass: Weighted mean of miscalibration factors in event i: For each ring, plot distribution of i, with each entry weighted by the fraction of reconstructed electron energy contained in the ring in that event Estimate of miscalibration for the ring given by position of peak obtained from Gaussian fit

35. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 35 Extraction of Calibration Coefficients Procedure is iterated until the calibration coefficients converge: Calibration coefficient for an individual crystal is the product of the ring coefficient C(  ) and the relative coefficient for the crystal within the ring obtained from Phi Independence:

36. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 36 Intercalibration of Crystal Rings Crystals within rings already intercalibrated using phi independence Event selection: require two “golden” electrons (i.e. little bremsstrahlung radiation), in order to minimize dependence on tracker material and hence Set up miscalibrations: 2% miscalibration between crystals within a ring (precision obtained after  -symmetry intercalibration) 5% miscalibration between rings Use simulated event sample corresponding to an integrated luminosity of 2.0fb -1 Intercalibration precision is RMS spread of residual miscalibration after correction:  = 0.6%

37. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 37 Intercalibration Precision Using Z  ee Intercalibration precision with 2 fb -1 for different ECAL barrel modules (increasing in  ): Intercalibration precision as function of event statistics: 2 fb -1

38. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 38 4: Inner Bremsstrahlung Photons in Z    

39. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 39 Z  : Introduction Radiative decays of Z bosons to muons: Clean source of high p T isolated photons with significant rate and very little background Energy scale determination independent of the ECAL Z  is an important tool for several commissioning tasks: Calibration between regions previously intercalibrated using W  e. Determination of overall energy scale, using Z mass constraint Probe for measuring photon trigger, reconstruction and identification efficiencies Tuning of algorithmic (clustering) energy scale corrections for photons and electrons (from photon conversions).

40. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 40 Z  : Preliminary Studies Investigated using events generated with full matrix element calculation of radiative decays Separate generation using ALPGEN and CompHEP Background: Z bosons produced with additional jet(s) Event selection: Muon pair invariant mass: 40 < M  < 80 GeV Reconstructed photon with p T > 15 GeV, within  R < 0.8 of either muon 3 body invariant mass: 87.2 < M  < 95.2 GeV Signal to background ratio: ~80 for 15 < E T  < 30 GeV ~1 photon per crystal for 1fb -1 of data

41. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 41 5. Low Mass Resonances

42. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 42 Calibration using  0   0  and  currently being studied as additional calibration tools Unconverted photons insensitive to tracker material distribution: No dependence on  Minimum separation for photons from  0 in the barrel with E T =5GeV is ~65mm 3 times crystal granularity For QCD events accepted by the Level-1 triggers, perform ECAL cluster reconstruction in small region identified by the trigger, using an online filter farm. Event selection: Require 2 ECAL clusters with energy in range 1.5 < E  < 5 GeV and separation at ECAL front face between 60 and 90mm. Tight requirements on ECAL shower shape to select unconverted photons Assuming a Level-1 trigger rate of 10 kHz: Rate after selection >1000  0 s/crystal per day of data taking at 2x10 33 cm -2 s -1 signal-to-background ratio ~2

43. Coseners House Forum on LHC Startup 13th April 2007 David Futyan Imperial College 43 Summary Accurate intercalibration of the CMS ECAL to the design goal of 0.5% essential for physics discovery reach, in particular for the H  channel. Can only be achieved in-situ using physics events. High p T isolated electrons produced from decay of W and Z bosons are key to achieving this goal. Baseline strategy: Phi independence of energy from jet trigger events to intercalibrate rapidly within rings at startup to 2% precision (< 1 day) Electrons from Z to intercalibrate between rings Once tracker is fully operational and aligned, design goal precision of 0.5% can be achieved with 5 fb -1 of data using single electrons from W  e