- comparisons we want to do for the Review (probably in the week of June 20th 2016) Baseline: FCal+ITK (LoI) mu=200 (range 190-210) and the others as additional comparison (mu=130-150, mu=70-90) - dedicated JES is currently being derived by Steve Alkire (also for sFCal and degraded FCal samples) - FCal gap width in simul: 267mu, 375mu and 500mu for FCal1,2,3, respectively to compare to: 1) sFCal(small gap)+ITK (LoI) with the same mu as FCal sample - sFCal gap width in simul: 100mu, 200mu and 300mu for sFCal1,2,3, respectively 2) degraded FCal+ITK (LoI), where degraded means that certain cells in FCal1 are switched off (0 energy) beyond a given |eta| cut - 7 variants are defined variant 1: (baseline) all FCal working 2: FCal1 |eta| > 4.25 dead 3: FCal1 |eta| > 3.65 dead 4: FCal1-3 |eta| > 4.25 dead 5: FCal1-3 |eta| > 3.65 dead 6: entire FCal1 dead 7: entire FCal1-3 dead -> production of DPD.pool.root files with self-made jets and with degraded FCal to be done by Namig & Sasha -> actually we produce xAODs with switched off cells, re-clustered topo-clusters, new jet collections and the towers from the reduced set of cells and clusters -> samples will be registered in some grid group area to not be lost (e.g. MPPMU) - these require dedicated JES for the variants 2-7 (i.e. 6 in addition to baseline) - variant 7 is probably easy to derive but it remains to be seen how MET reacts to this ... - priority should be on the variants 2,3 and 6 (in this order) since jets are not used in regions without any response. - to produce LCTopo and EMTopo jets (MyAntiKt4LCTopoJets and MyAntiKt4EMTopoJets) without any calibration use reco.py from: /afs/cern.ch/user/m/menke/public/xAOD_sFCal_JetRec/ - in principle LC calibration could also be re-done but we keep the corrections to the jet level since on cluster-level most em-deposits are entirely missing with FCal1 dead ... - what physics/performance studies are achievable in time for the review + single and double tag efficiency for forward jets in VBF samples for FCal/sFCal -> do we need the final JES from Steve for this? -> no, the idea is to use constituent level jet and compare ad-hoc PileUp suppression techniques, like jet width, PTD, soft-killer, self made jet-area subtractions based on (tower-based or cluster-based) rho - median rho distribution in slices of eta made from positive clusters or towers with a width of Delta eta ~= 0.5 and sliding in steps of 0.1 in eta ( rho = pT/area, whre area is 0.1x0.1 for towers and for clusters area=? for clusters: run FastJet jet maker on all positive clusters and make pseudo jets out of them remember the pointer to cluster and let FastJet calculate the area for each constituent (AntiKt4 jet algo) -> this is then the area of the cluster) -> use these rho-definitions to apply area based corrections to constituent level jets in FCal/sFCal samples and compare efficiency as function of pT + signal and PileUp jet reco efficiency after applying some grooming on the AntiKt4 jets (e.g. soft-killer which acts on constituents) in VBF samples -> see generator level study from Peter where grooming (jet area subtraction) is suppressing PileUp in the forward by a factor ~1000 without significant signal loss (plots on slide 13 of "VBF jets" attached to Monday) + quark/gluon separation in forward jets (related to PileUp suppression in terms of jet moments and jet sub-structure) + have cross-sections for VBF heavy and SM Higgses handy and their expected event yields for 3000 fb^-1 @ 14 TeV + slides from Doug Schaefer (attached to Wednesday) show that for a SM Higgs the impact of PU rejection in the FCal region improves signal significance by up to 28% with 5 times better PU rejection. -> but SM Higgs is not the best laboratory for forward jets -> impact is larger on heavy Higgs where more tag-jets go forward + Z->ee samples with one forward electron and one central electron (look at quick plot from slide 7 of Z ee contribution added to Tuesday) -> compare Z-mass and reco efficiency in FCal/sFCal/degraded FCal -> electron fakes in the forward from PileUp + impact on MET is coming mainly due to the jets and their PileUp suppression - there is no PileUp suppression yet for soft term -> the overall impact on MET will likely be small - granularity will not change in 2.5 < |eta| < 3.2 which is coarse - for some comparison of MET impact could use just hard objects (jets) and the pT fraction covered (or lost) by the FCal/sFCal/degraded FCal -> this can be calculated directly from jets (no need for MET framework) -> we don't need the soft term for this - look into the samples with large MET (>100 GeV) (for example the ttbar sample with at least one lepton and MET > 100 GeV mc15_14TeV.407099.PowhegPythia8EvtGen_A14_ttbar_hdamp172p5_MET100) + W's in the forward from Vector Boson Scattering are interesting from the theory point of view but: - boson tagging in the forward would be a complete new development - the x-section of these forward boosted W's is small - maybe the boosted W's from the tt-bar sample are candidates to study fat W jets in the FCal/sFCal + Delta R for reco and truth jets as function of eta/pt for sFCal/FCal/degraded FCal to study position resolution effects + Studies with the standalone simulations of FCal/sFCal - Bengt focused on timing and the reflection of signals at the rod ends to get the intrinsic timing resolution of the detector - reflections are the main cause of limits to the time resolution - the smaller cells in the sFCal will cause less PileUp and signal dilution and thus allow for better timing resolution but the intrinsic properties (reflection at the rods) is similar for FCal and sFCal - the drift time in the smaller gaps of the sFCal will be shorter and this makes the signal faster (by a factor 267/100?) and could impact the timing resolution - a ~60 ps time resolution could help to suppress PileUp offline (and HLT) - single pion/electron position resolution from two PhD students from Morocco + task list 1) finding the signal - jets }___\ - PU suppression technique is needed - electrons } / + jet area (Tom) + soft-killer (Tom) jet area: - inputs: clusters E >0 or towers E > 0, jets can use either one (EM scale for towers, LC scale for clusters) - density measure rho (needs no final JES)) - three different approaches for rho/jet: a) rho_EM from towers for EMTopo jets b) rho_LC from clusters for LCTopo jets c) rho_EM from towers scaled to LC-scale for LCTopo jets - area calculation for rho: + TopoCluster: Voronoi area from FastJet + Towers: constant area 0.1x0.1 or Voronoi area (check that both are compatible) - rho(eta) = median(pT/area) (take median of individual rho in eta regions per event) - jet correction: pT_jet^cor = pT_jet^constituent - A_jet*rho(eta_jet) soft-killer: - inputs same as for rho (E>0) - filter inputs in "quadratic" patches in eta x phi (e.q. 0.5 x 0.5) by removing all inputs inside each patch below the median of the patch - use filtered list as input to jets - get EMTopo and LCTopo jets from filtered inputs -> no additional calibration applied - in principle a final JES can be applied to these (but might not be feasible in time) - need recipes how to run these in RootCore, athena 2) tagging and scoring - truth association + Delta R matching for truth and reco jet (pT dependency? 0.4?) -> distribution to optimize Delta R cut to be done by Olga + truth label of jets (quark/gluon) -> recipe on how to get the truth parton label to be provided by Olga + truth label for electrons and forward electrons? -> to be looked at by Olga - scoring + efficiency for signal & background - for event filtering - for individual objects before/after PU suppression (no JES needed) -> make sure consistent normalization is used (what is 100%? e.g. decay products of H + >1 jet) - efficiency of forward jets as function of eta coverage and pT cut -> VBF samples to make simple cut based selection and provide scoring recipe (Sven + Peter) - Z->ee with one forward electron reco efficiency and mass resolution (Olga) -> fake rate of forward electrons (from PU jets) -> m_Z^reco with and without truth label in mass window of +/- 30 GeV around Z^0 mass - MET (Teresa) -> (tt-bar with >0 leptons & > 100 GeV MET on generator level) effect of jets in FCal on MET start with LCTopo jets on constit scale and refine with JES later compare before/after PU suppression 3) VBF standard analysis comparison on FCal/sFCal geometries (Ilya eta al.) 4) refinement (lower priority) - quark/gluon tagging in PU suppression PU is gluons and stochastic jets -> gluons are wide jets -> stochastic jets are wide and flat VBF jets are quarks -> quark jets are narrow jet width w = 1/pT * sum (pT_^alpha * Delta R_i^beta), (assume alpha = beta = 1 case) PTD = sqrt(sum pT^2) / sum(pT) -> study both on jet-level using jet constituents -> jet and width as function of eta and pT for FCal/sFCal (Denis) 5) position and energy resolution of single particles - smaller gaps degrade intrinsic energy resolution (for 100 GeV electrons the resolution is 7.5% worse for the sFCal without PileUp -> provide final numbers from stand-alone simulations (for FCal/sFCal geometries) (Bengt + students) - finer granularity improves position resolution -> from Bengt's studies by a factor ~2 in phi and a smaller factor in eta - could compare to electrons with PileUp from Z->ee and/or ttbar (Olga for Z->ee) 6) noise in the FCal region for FCal/sFCal with large mu (see plot in scoping document) -> all cells with finer granularity the noise is a factor 2.5 smaller in the sFCal (4 would be the factor if all small cells making one large cell are fully correlated and sqrt(4)=2 would be the factor if they are totally uncorrelated. The observed 2.5 mean we gain information by increasing granularity) 7) energy and position resolution of jets -> from di-jet samples with and without final JES and with and without PU suppression (Sven) (gluon/quark fraction? difference?) 8) mention possible impact on connection with ITK by improved position resolution. -> this is mainly important for electrons and not so much for jets 9) track-based PielUp suppression for FCal/sFCal (studies by Marianna). -> is the tracking improving both FCal and sFCal without discriminating? -> or can we get better performance from either geometry in connection with ITK? 10) for HLT same arguments as for offline apply -> i.e. increased granularity improves PU suppression for L1 (L0) the improvement depends on the implementation of SuperCells. Potentially their granularity can be increased but the current outline of SuperCells is even coarser than FCal1 granularity. Timeline week of June 20th : The Review (reviewers are picked by ATLAS management (Kevin and Rob)) -> full day event with different sessions on hardware/risks/physics/performance June LAr week (week of June 6th): Presentation of the material for the review May (bi-weekly sFCal analysis meetings) use status reports in these to finalize material for the review -> May 12 to present status April 28 (bi-weekly sFCal analysis meeting) first status reports about assigned tasks -> do we have any technical show stoppers? -> also discuss this in the time before via e-mail