1. On behalf of the ATLAS MET group Ellie Dobson e.dobson1@physics.ox.ac.uk Missing transverse energy (MET) performance in ATLAS Combined performance session ATLAS UK Durham meeting 09-01-08  MET Reconstruction  Fake MET  MET in early data  Min Bias  Z→ll events  Z →  events  W→μν events  Semileptonic ttbar events

2. 2 MET in ATLAS A good measurement of MET is critical for the study of many physics channels in ATLAS especially for the search for signals from new physics such as SUSY or extra dimensions. It is also necessary for precision studies such as a W mass measurement and for the reconstruction of the Higgs boson mass in H→ . CSC Authors S.Asai, E. Barberio, D. Casadei, D. Cavalli, X.Chen, M.Consonni, L.Courneyea, K. Cranmer, R. Djilkibaev, E. Dobson, R. Duxfield, L.Flores, A. Gibson, A. Gupta, N. Kanaya, E. van der Kraaij, Y.Ishizawa, R. Lafaye, R.McPherson, B.Mellado, S. Menke, A. Mincer, H.Okawa, S.Padhi, F. Paige, A. Phan, C. Pizio, J. Poveda, R.Prabhu, S.Resconi, M. Rijpstra, G. Rosenbaum, A. Schwartzman, A. Shibata, R. Teuscher, D. Tovey, I.Trigger, G. Usai, M.Vreeswijk, L. Zhao, S.L.Wu, S.Yamamoto, A.Yurkewicz Further details available on the CSC note draft 3 which is ready and available on https://twiki.cern.ch/twiki/bin/view/Atlas/EtMissCSCNote

3. 3 MET reconstruction MET_CalibMET_CryoMET_Muon ++ - Sum over calorimeter cells surviving noise cuts at Cell energy > 2σ(noise) - Map filled between cells and associated objects (electrons, photons, taus, jets) so no double counting (Refined calibration) - Calibration weight assigned according to the reconstructed object type (H1 like calibration for jets) - Cells outside of objects also included - Correction for energy lost in the crystat between the EM and hadronic calorimeters - Reconsutructed MuonBoy muons used - Energy lost by muons in calorimeter already accounted for so no double counting - Primary motivation to reliably reconstruct SM Higgs mass from H- tt decay channel. Main contributor to this mass resolution is the the MET resolution. Need quiet final states so low pt depositions (pions, pile up) are more relevant. - Classification not so important for (example) SUSY (with hard jets and large MET). - Computed MET using clusters from objects, not cells Object based reconstruction Cell based reconstruction

4. 4 Evaluation of MET performance Resolution = σ ( MET (RefFinal)- MET (truth) ~ Displacement at very high regions (high pt jets) and very low regions (noise suppression method) in low and medium regions of SumEt Tails (fraction of events outside 3σ of distribution) Angular resolution Dependence on the event topology Linearity = (MET (truth)- MET (RefFinal))/MET (truth)

5. 5 Fake MET : MET (rec) – MET (true) SourcesMethod of removal Cosmic rays- Use calorimeter timing to compute ‘up-minus-down’ time Beam related background- Use calorimeter timing to identify out of time events Muon mismeasurements - Inefficiency in reconstruction - Reconstructing fake muons (mainly punch through from jets) - Look for events with large depositions in outermost calorimeter layers to identify punch through - Compare track parameters with reconstructed quantities to identify discrepancies Calorimeter mismeasurement - Hot/dead/noisy cells - Jets in dead regions - Algorithms to veto events with large fake MET - Cut on high Pt jets close to MET vector - Remove jets pointing to dead regions

6. 6 MET in early data: Minimum bias events Soft interactions of the beams Dominated by low Pt events - don’t have large energy imbalance or MET Useful for comissioning in early runing due to large statistics and less complicated event selection MinBias J0

7. 7 MET in early data: Z→ll events x y e+ e- Bisector 2 (parallel) Bisector 1 (perpendicular)  This choice of axis leads to an optimal MET resolution Resolution axis = Bisector 2 if angle between electrons > π/2 Resolution axis = Bisector 1 if angle between electrons < π/2

8. 8 e+ e- Hadronic recoil Axis EtMiss Y axis quantity (red) negative == EtMiss anti parallel to axis == Magnitude of hadronic activity underestimated Resolution axis Perpendicular to resolution axis MET in early data: Z→ll events : scale

9. 9 MET in early data: Z→ll events : Resolution

10. 10 MET in early data: Z →  → jet l events Mass reconstruction possible due to the collinear approximation (directions of the two neutrino systems from each  decay are coincident with the ones of the measured decay products) Mass reconstruction is dominated by the MET measurement Can determine MET scale from the peak position of the  invariant mass distribution Opposite Sign (OS) lepton-hadron for signal. In backgrounds same probability to have OS or Same Sign (SS) lepton-hadron: evaluate background in situ from SS events.

11. 11 MET in early data: W→lν events - Comparing ratio R between neutrino and lepton pt: (sensitive to the scale but not to the resolution of the MET) - Template method: Convolute true transverse W mass distribution with MET response to create a set of template histograms with which to fit the transverse mass distribution:

12. 12 MET in early data: Semileptonic ttbar events - W transverse mass Effect of fake MET Effect of MET scale

13. 13 MET in early data: Semileptonic ttbar events – kinematic fit Compute chi squared values on a fit to the known kinematical variables in the events - Mw(had) = 80.4 GeV - Mw(lep) = 80.4 GeV - Mtop(lep) = Mtop(had) - Mtop (had) = 175 GeV Fit is dependent on MET- take MET scale that gives the lowest chi squared value

14. 14 Conclusions MET reconstruction techniques now relatively established Emphasis now on developing methods to check and improve MET performance in early SM data without relying on MC truth What we learn from early SM data may be applied to new physics searches in the future