Exploring Physics and Detection Strategies at the ATLAS Workshop in Les Houches, 2003

1. The first 10 fb-1 in ATLAS Les Houches Workshop, 27/5/2003 Fabiola Gianotti (CERN) Giacomo Polesello (INFN Pavia) • Which physics ? • Which detector ? • Which data samples for physics and calibration ?

2. Process Events/sEvents for 10 fb-1Total statistics collected at previous machines by 2007 W e 15 108 104 LEP / 107 Tevatron Z ee 1.5107107 LEP 1107104 Tevatron 1061012–1013109 Belle/BaBar ? H m=130 GeV 0.02105 ? m= 1 TeV 0.001104--- Black holes 0.0001103--- m > 3 TeV (MD=3 TeV, n=4) Expected event rates at production in ATLAS at L = 1033 cm-2 s-1 • Already in first year, large statistics expected from: • -- known SM processes  understand detector and physics at s = 14 TeV • -- several New Physics scenarios • ~ 107 events to tape every 3 days assuming 30% data taking efficiency •  statistical errors negligibleafter few days

3. Understand and calibrate detector and trigger in situ using well-known physics samples e.g. - Z  ee,   easy-to-trigger channels - tt  b bjj Also useful to debug/optimize software (reconstruction, …) • Look for New Physicspotentially accessible in first year : • e.g. -- SM Higgs • -- SUSY (squarks, gluinos, Higgs bosons, ….) • -- others …. including surprises ? over some mass ranges Which studies and physics the first year ? • Understand basic SM physics at s = 14 TeV first checks of Monte Carlos • e.g. - measure cross-sections for e.g. minimum bias, W, Z, tt, QCD jets (to ~ 10-20 %), • events features, particle multiplicities, pT and mass spectra, angular distributions, etc. • - measure top mass (to 5-7 GeV)  give feedback on detector performance (jet scale …) • Prepare the road to discovery: • -- measure backgrounds to New Physics : e.g. tt and W/Z+ jets (omnipresent …) • -- look at specific “control samples” (background-enriched) for the individual channels: • e.g. ttjj “gauges” ttbb background to ttH Note:if mH 120 GeV : fast Higgs discovery may be crucial in case of competition with Tevatron May be most difficult physics goal for first year …

4. The first “physics data” to commission the detector : cosmics • End 2006- beg 2007: 1-3 months of data • taking with cosmics. • Full simulations of cosmic muons in ATLAS • started (include cavern overburden) • Expected rates in the detector : few Hz •  very useful events for debugging the • detector, first tracker alignment, • calorimeter energy reconstruction, etc.

5. A typical event …. • One track reconstruted in Muon chambers • Two tracks reconstructed in Inner Detector

6. Same event: longitudinal view

7. April-May 2007 : only one beam in the machine (most of the time …) • here “physics data” are beam-halo muons and beam-gas interactions. Expected in ATLAS: -- ~ 105 (108) beam-halo muons in tracker (muon chambers) -- ~ 106 tracks in tracker from beam-gas

8.  Well-known, clean processes from standard trigger menu: e.g. • Z  ee : ECAL inter-calibration, absolute E-scale to ~ 0.1%, etc. • Z   : p-scale in tracker and Muon Spectrometer, etc. • tt  b bjj : absolute jet-scale from W  jj (~1%), study b-tag, • reconstruction of complex final states (for ttH), etc. ~ 6 x 104 evts/day after cuts ~ 104 evts/day after cuts, S/B ~ 65  Additional lower-thresholds samples (pre-scaled triggers) : • Minimum-bias events :pp interaction properties, MC tuning, LVL1 efficiency, • radiation background in Muon chambers, etc. • QCD jets (20 ET400 GeV): QCD cross-sections and MC tuning, • trigger efficiency, calorimeter inter-calibration, • jet algorithms, background to Higgs, SUSY, etc. • Inclusive e pT > 10 GeV: trigger efficiency, ECAL calibration, ID alignment, • E/p, e reconstruction at low-pT, etc. • Inclusive  pT > 6 GeV: trigger efficiency,  reconstruction at low-pT, • E-loss in calorimeters, ID alignment, etc. These are only few examples … ~ 107 events per sample needed  10% of total trigger rate under normal operation (more at the beginning) First pp collisions : collect data for calibration and to understand “basic” physics

9. here discovery “easy” with H  4 mH > 114.4 GeV H  ttH  ttbbqqH  qq ( + -had) S 13015~ 10 B 4300 45~ 10 S/ B2.02.2~ 2.7 total S/ B SM Higgs Then : the “discovery phase” can start ? mH ~ 115 GeV 10 fb-1

10. H   ttH  tt bb  b bjj bb qqH  qq b b  H  Remarks: Each channel contributes ~ 2 to total significance  observation of all channels important to extract convincing signal in first year(s) The 3 channels are complementary robustness: • different production and decay modes • different backgrounds • different detector/performance requirements: • -- ECAL crucial for H   (in particular response uniformity) : /m ~ 1% needed • -- b-tagging crucial for ttH : 4 b-tagged jets needed to reduce combinatorics • -- efficient jet reconstruction over || < 5 crucial for qqH  qq : • forward jet tag and central jet veto needed against background Note : all require “low” trigger thresholds. E.g. ttH analysis cuts : pT () > 20 GeV, pT (jets) > 15-30 GeV

11. H  qqH  qqH  4qqH  qqWW ( + -had) S 120~ 8~ 518 B 3400~ 6< 115 S/ B2.0~ 2.72.83.9 mH ~ 130 GeV 10 fb-1 = e, total S/B  6 • 4 complementary channels for physics and for detector requirements • S/B < 3 per channel (except qqWW counting channel) observation of all channels • important in first year • H  4 low rate but very clean: small background, narrow mass peak • Detector requirements: • --  90% e,  efficiency at low pT (analysis cuts : pT 1,2,3,4 > 20, 20, 7, 7, GeV) •  in particular low di-lepton LVL1 thresholds • --  /m ~ 1%, tails < 10%  E, p measurement and resolution in ECAL and tracker at low pT

12. A/H  , tg  = 38 m~11 GeV MSSM Higgs bosons h, H, A, H  mh < 135 GeV mA mH mH at large mA 5 discovery curves A, H, H cross-section ~ tg2 Best sensitivity from A/H  , H   bbA/H   : -- covers good part of region not excluded by LEP -- experimentally easier than A/H   -- crucial detector :Muon Spectrometer (high-pT muons from narrow resonance)

13. Here 5 discovery of bbA/H  4b possible at Tevatron with 15 fb-1

15. Large cross-section   100 events/dayat 1033 for • Multijet + ETmiss is most powerful and model-independent signature (if R-parity conserved) ATLAS 5 discovery curves ~ 100 days : up to 2.3 TeV ~ “10 days” : up to 2 TeV ~ “ 1 day” : up to 1.5 TeV SUPERSYMMETRY Signal could be confirmed by (more model-dependent) lepton signatures

16. Peak position correlated to MSUSY Events for 10 fb-1 background signal Events for 10 fb-1  Tevatron reach ATLFAST ET(j1) > 80 GeV ETmiss > 80 GeV signal background From Meff peak, first/fast measurement of SUSY mass scale to  20%(10 fb-1, mSUGRA) Detector/performance requirements: -- calorimeter coverage and hermeticity for||<5 -- calorimeter energy scale calibration to ~5% -- “low” Jet+ETmiss trigger thresholds for low masses at overlap with Tevatron region (~400 GeV)

17. reconstructed Events with ETmiss > 50 GeV if leading jet undetected these 2 events contain a high-pT neutrino • Cracks : • -- can be monitored with Z ( ) + jets • -- impact minimised by ETmiss isolation, removal of jets in cracks Z ( ) + jet full simulation • “Poor” initial calorimeter calibration may increase trigger rates  impact on low-mass SUSY (Very) pessimistic uncorrected non-compensation simulated by + 20% enhancement of EM scale + 50% rate for ETmiss > 80 GeV

18. Which detector the first year ? staged staged in part staged staged Staged detector components: -- 1 pixel layer -- TRT outer end-cap -- Gap scintillator -- EEL/EES MDT and half CSC -- Part of forward shielding -- Part of LAr ROD -- Large part of HLT/DAQ processors staged Guiding physics principles: -- all sub-detectors needed already in 1st year -- physics potential decreases fast with decreasing h coverage (e.g. H   significance decreases linearly) -- full radial redundancy in tracking less crucial at ~ 1033  Technical (e.g. installation) and schedule constraints

19. Staged items Main impact during Effect first run on 1 pixel layerttH  ttbb~8% loss in significance Gap scintillatorH  4e~8% loss in significance MDTA/H  2~5% loss in significance for m~ 300 GeV Trigger processorsB-physicsprogram jeopardised High-pT physicsno safety margin (e.g. for EM triggers) Requires 10-15% more integrated luminosity to compensate. Complete detector needed at high luminosity: -- robust pattern recognition (efficiency, fakes rate) in the presence of pile-up and radiation background -- muon measurement -- powerful b-tag -- robustness against detector aging and L > 1034 -- precise measurements (e.g. light Higgs) may require low trigger thresholds at (very) high pT Summary of physics impact of staging initial detector

20. Conclusions • ATLAS has potentially an impressive physics programme right from the beginning. • Event statistics : 1 day at LHC  10 years at previous machines in some cases • Construction quality checks and beam tests of series detector modules show that the • detector “as built” should give a good starting-point performance. • However,a lot of data (and time …) will be neededat the beginning to: • --commission (to < 1% in most cases) the detector and trigger in situ • --reach the performance needed to optimise the physics potential • --understand “basic” physics at s = 14 TeVand normalise MC generators • --measure backgrounds to New Physicsextract a convincing “early” signal • Redundancy from several control samples is mandatory, especially for difficult channels • (e.g. light Higgs) and measurements (e.g. W mass) • The initial “physics data”(cosmics, beam-halo, beam-gas, first collisions, etc.) are • therefore crucial to reach quickly the “discovery-mode”.

21. --HLT/DAQ deferrals limit available networking and computing for HLT  limit LVL1 output rate -- Large uncertainties on LVL1 affordable rate vs money (component cost, software performance, etc.) Selections (examples …) LVL1 rate (kHz) LVL1 rate (kHz)LVL1 rate (kHz) L= 1 x 1033 L= 2 x 1033L= 2 x 1033 Real thresholds set for no deferrals no deferrals with deferrals 95% efficiency at these ETAn example for illustration… MU6,8,20 23 19 0.8 2MU6 --- 0.20.2 EM20i,25,25 11 12 12 2EM15i,15,15 24 4 J180,200,200 0.20.2 0.2 3J75,90,90 0.20.2 0.2 4J55,65,65 0.20.2 0.2 J50+xE50,60,60 0.40.4 0.4 TAU20,25,25 +xE30 2 2 2 MU10+EM15i --- 0.10.1 Others (pre-scaled, etc.) 5 55 Total ~ 44 ~ 43~ 25 LVL1 designed for 75 kHz  room for factor ~ 2 safety Likely max affordable rate, no room for safety factor

22. Selection (examples …) Rate to storageat 2x1033 (Hz) Physics motivations (examples …) e25i, 2e15i ~ 40(55% W/b/c  eX)Low-mass Higgs (ttH, H 4, qq) 20i, 210 ~ 40(85% W/b/c  X) W, Z, top, New Physics ? 60i, 220i ~ 40(57% prompt )H  , New Physics (e.g. X   yy mX~ 500 GeV ) ? j400, 3j165, 4j110 ~ 25 Overlap with Tevatron for new X  jj in danger … j70 + xE70 ~ 20 SUSY : ~ 400 GeV squarks/gluinos t35 + xE45 ~ 5 MSSM Higgs, New Physics (3rd family !) ? More difficult high L 2m6 (+ mB )~ 10 Rare decays B  X Others ~ 20 Only 10% of total ! (pre-scaled, exclusive, …) Total ~ 200 No safety factor included. “Signal” (W, , etc.) : ~ 100 Hz Best use of spare capacity when L < 2 x 1033 being investigated Which data samples ? Total trigger rate to storage at 2 x 1033 reduced from ~ 540 Hz (HLT/DAQ TP, 2000) to ~ 200 Hz (now) High-Level-Trigger output

23. Assuming : • 109 events recorded in first year • Raw data event size : 1.6 MB • Reconstructed event size : 0.5 MB • Time to reconstruct 1 event : 640 SI95-sec -- was 2.2 MB --zero-suppression in calorimeters not included yet Storage for raw data 1.6 PB(back-up not included) Storage for reconstructed data (ESD) 1.0 PB(present + previous version) CPU for reconstruction 178 kSI95 (1st pass + calibration + 2 re-processings) Implications on Computing at CERN Tier-0 functionality

24. 448 channels in total Source Expected contribution to cL (over  x  = 0.2 x 0.4) Geometry (e.g. residual Accordion modulation) 0.25-0.35 % Mechanics (absorber and gap thickness) < 0.25% Calibration (amplitude uniformity, ~ 0.4 % difference physics-calibration) Total 0.5-0.6% Expected “realistic” initial performance and calibration strategy One example : EM calorimeter H   : constant term ctot  0.7% over || < 2.5needed to observe signal peak on top of huge  background Strategy to achieve this goal  By construction (e.g. mechanical tolerances): expect cL ( “local” constant term)  0.5% over  x  =0.2 x 0.4 There are ~ 400 such regions in || < 2.5

25.  Beam tests of 4 (out of 32) barrel modules and 3 (out of 16) end-cap modules in 2001-2002: from first results step  is achieved • 1 barrel module: •  x  = 1.4 x 0.4 • ~ 3000 channels  Scan of a barrel module with 245 GeV e-  over ~ half module r.m.s.  0.67% over ~ 500 spots “On-line” uniformity : ~ 1.3% Uniformity after corrections (e.g. optimal filtering) :~ 0.67 % Uniformity after corrections over  x  = 0.2 x 0.4 :~ 0.55%

26. rate ~ 1 Hz, ~ no background, allows standalone ECAL calibration • Nevertheless, let’s consider the worst (unrealistic ?) scenario : no corrections applied • cL = 1.3 % measured “on-line” non-uniformity of individual modules • cLR = 1.5 % no calibration with Z  ee ~ 250 e per region needed to achieve cLR  0.4% ctot = 0.5%  0.4%  0.7% ~ 105 Z  ee events (few days of data taking at 2 x 1033) ctot 2% conservative : implies very poor knowledge of upstream material (to factor ~2) H   significance mH~ 115 GeV degraded by ~ 25%  In situ calibration with Z  ee events ctot = cL cLR cLR long-range response non-uniformities of the 400 regions (module-to-module variations, different upstream material, etc.) (from full simulation)