Electron/Photon group overview

US CMS Meeting, Princeton

April 30, 2004

Rick Wilkinson, Caltech

USCMS in e/gamma

• UCSD: Higgs – Jim Branson, Satyaki Bhattacharya, James Letts, Kyle Armour

• Caltech: Higgs calibration– Harvey Newman, Sergey Shevchenko, Vladimir Litvin, Tony Lee

• Caltech: Calorimetry core software– Vladimir Litvin, Rick Wilkinson

• Yale: Calorimetry core software– Homer Neal

• Minnesota: Simulation– Maria Obertino

• UC Riverside: e/gamma software, calibration– David Futyan

Higgs (Caltech) • Traditional, cuts-based counting experiment• Background simulation uses generator-level preselection

– Look for e’– Saves factor of ~3000 in CPU for QCD background

• Resulting luminosity required 5discovery:– Inclusive Higgs production (pp H ) 39.2 fb-1

• Also look for Vector Boson Fusion– pp qqH qq– Has two forward jets with || ~ 3

– Surprisingly good discovery reach, 41 fb-1

Higgs (UCSD) • More aggressive; avoid cuts. Keep all the information you can.• Sort events by their cleanliness, using

– Photon quality (narrowness of the worst one)

– Kinematics, using a neural net

– Even use the lineshape of the Higgs mass hypothesis!

• Combine all these factors into a S/B estimate for the event• Plot the event by its S/B

• Results are amazingly good!– 5 discovery only needs:

• 2 fb-1 for jet-jet bg

• 2 fb-1 for -jet bg

• 0.5 fb-1 for bg

– Need to combine somehow– Too good? log(s/b)

background

signal

H photon quality• Categorize events by the quality of

their worst photon.

• r9 = (Sum of 9)/ESC (uncorrected)

– 4 bins in narrowness r9

x2 bins (barrel, endcap)

makes 8 categories of events

• Better photons have– better mass resolution

– Less QCD background

• Analyze event categories separately– Only combine in final plot

signal

unconverted

background

H Kinematics Neural Net

• Neural Net Inputs are:– Jet-jet and -jet

• Calo isolation, track isolation, ET1/(ET1+ET2), ET2, |1-2|

– Irreducible background

• ET1, ET2, ESC1, ESC2, |1-2|

• S/B obtained from the black fitted curves

γγ-jet cat1 -jet cat1 (cleanest)(cleanest)γγ-jet cat1 -jet cat1 (cleanest)(cleanest)

background

signal

H ass shape & discovery reach

• Include mass information in s/b• Fit resulting plot for signal,

background

• Do many trials:– background-only experiments

– signal+background.

• Some overlap– Luck will play a role in how

fast we find the Higgs

log(s/b)

backgroundsignal

Calibration• Baseline: Track momenta from electrons from W decay

– Problem: Can we avoid strict cuts on brem?– May take months

• Other, faster techniques– (V. Litvin & S. Shevchenko, Caltech)

• Photons usually separated by 3-10 crystals

• Needs a day or two of dedicated running with the full DAQ bandwidth!

– Combine -symmetry + Z ee• See next slide

Calibration• Start by looking at -symmetry, comparing summed

energy in crystals around a ring in D. Futyan, UCR)– In min-bias events

• Too low energy?• Sensitive to tracker material

– In jet triggers• Trigger biases

– trigger region boundaries!

• Then, calibrate between the rings with Z ee (Rome)– 170 parameters in barrel, 80 in endcap– Math. Lots of math. (Iterative algorithm now, others possible)

All rings combined

ETvs

Calorimetry Software

• Skeleton transplant in progress!– Switching to common framework with Tracker, Muon– Allows us gain functionality they already have:

• Track propagation• DAQ readout grouping• Misalignment

• To-do list:– Calibration constants– Analyze HCAL testbeam data with ORCA

e/gamma code: Physics ObjectsPersistent Physics Objects in DST data for Data Challenge ’04 datasets:(D. Futyan, E. Meschi)• EGBCluster (basic cluster)• ET threshold gives EGCluster• Brem recovery gives EGSCluster (supercluster)• Endcap preshower gives EGECluster• Fiducial cuts give EgammaCandidates

– Offline • EgCandFromEGSCluster• EgCandFromEGECluster

– Level 2 trigger• EgCandL2FromEGSCluster• EgCandL2FromEGECluster

• If there’s an associated pixel track: EGElectron• If no associated pixel track, EGPhoton• Also EPTrack, EgammaMC

GEANT4/OSCAR Validation

• Long-running mystery about the electron energy resolution.

• There was a bug in the simulation thresholds used in the material description. Tracker cooling ledges were opaque to their own brem. (M. Obertino)

CMSIMOSCAR

Emeas/Etrue

Why this differencein the energy

distribution ?

Electrons: Resolution vs.

• ECAL Barrel resolution gets worse with TakahashiICL) Doesn’t seem to be because of lateral shower spread.

• Maybe back or front leakage?

<2nd sub-module> <16th sub-module>

back leakage

front leakage

~26 radn.lengths

<<26 radn.lengths

High Energy Electrons• For Randall-Sundrum graviton studies (Collard, Lemaire)• Need to re-optimize clustering algorithms & corrections• Synchrotron radiation not a problem• ADC saturation is a problem, but can be corrected