Update on the Diphoton + MET Analysis Bruce Schumm, channeling Ben Auerbach (Argonne), Osamu Jinnouchi (Tokyo Tech), Susan Fowler (Penn) UC Santa Cruz / SCIPP 6 March 2014 SUSY General Analysis Meeting
06 March Conference Note public in early January: Search for Supersymmetry in Diphoton Events with Large Missing Transverse Momentum in 8 TeV pp Collision Data with the ATLAS Detector ATLAS-CONF List of Tasks generated to move from note to paper…
04 June First: MC performance on tight-tight sample LocHadTopo has slightly larger tails than MetRefFinal
04 June QCDtg+Iso close to tight-tight (signal) distribution proxy for high MET QCDtg provides good representation of tight-tight MET distribution QCDg+Iso also looks good but statistics are low.
04 June For LocHadTopo, both QCDtg and QCDg seen to provide a good representation of the tight-tight MET distribution (again, using QCDtg-Iso as a proxy at high MET)
16 July Next: Signal Regions We define five signal regions, for: Strong production, high and low bino mass (SP1,SP2) Weak production, high and low bino mass (WP1,WP2) Choose MET cut to suppress backgrounds (MIS)
04 June Direct Background Estimate Methodology ACB e.g. for signal region WP2 Estimate = C*(A/B) Nominal control sample is QCDtg_50_noIso This is METRefFinal; can also look at QCDtg and QCDg of LocHadTopo MetRefFinal
19 December Nominal QCD Background Control Region Study
19 December We have performed the WP2 QCD background estimate without the \dphij cut, finding that the expected background rises from $0.90 \pm 0.35$ to $1.7 \pm 0.5$. It should be noted that the background would be expected to rise with the removal of the \dphij cut; if the `gg' distribution of Fig.~\ref{fig:dphij} is the correct distribution of the WP2 QCD background, this increase would be about 20\%, to 1.1 events. The observed value of 1.7 events is approximately 50% higher than this, which we interpret as an additional 50% systematic uncertainty on both the WP2 and MIS QCD background estimates. DPHI_JET_MET SYSTEMATIC
19 December QCDg+Iso Comparison LocHadTopo Comparison
16 July SP1 M eff Extrapolations
16 July SP2 M eff Extrapolations
16 July Combining all the above information yields the following overall result for QCD background (See Note for justifications…)
04 June Electroweak Backgrounds (W , ttbar, etc.) ~75% involve e fake; much of remainder incorporated in QCD backgrounds Reconstruct e sample; scale by measured e fake rate +/- 25% uncertainty from non e fake processes +/- 10% uncertainty from fake rate measurement
04 June e Fake Rate Results
04 June e Sample Statistics and EW Background Estimates
04 June Irreducible Backgrounds Z ; Z Small contribution NLO K-factor 2.0 +/- 0.3 (well understood) Estimate directly from MC W ; W l Larger contribution NLO K-factor 3.0 +/- 3.0 Dominant background systematic Constrain with new data-driven study
04 June W K Factor LO “radiation zero” eliminated at NLO Grows with hardness of radiation Rapidly-varying function of W system recoil
04 June Can we constrain the W K factor with an lgg (l = e, ) sample? Ben Auerbach
04 June Choose study region to be 50 < MET < 250 (leave MIS signal region blind) P T (l ) > 100 Ben Auerbach N expected = 7.4 (6.5 W ) N observed = 7.0 K factor of 3.0 +\- 1.2
16 July Irreducible background results And then putting it all together…
16 July < MET < 100 Sideband Studies – No Dphi cut 100 < MET < 150 Sideband Studies – No Dphi cut
16 July
16 July Wrap-Up Preliminary estimates of background completed Note updated with new background studies (nearly done) ATL-COM-PHYS Addressing comments from prior review (before p1328/p1181 MET changes that threw us back) Starting to build toward request for unblinding In the mean time, are developing limit-setting approach, and beginning to evaluate signal systematics
04 June MET Issues Are latest (“post-Moriond”?) object definitions included in p3128 EGamma10NoTauLoose MET? We will need in any case to assemble our own “fluctuated” EGamma10NoTauLoose in order to do systematic studies But for now, background estimates largely insensitive to MET systematics (data-driven), so could use intrinsic p1328 variable if “approved” Will definitely need to be able to assemble EGamma10NoTauLoose from scratch soon though.