Alan Watson L1Calo Upgrade Meeting 1 EM Rejection in Phase1 Developments since Stockholm: Using depth information aloneUsing depth information alone Using.

Slides:



Advertisements
Similar presentations
STAR Status of J/  Trigger Simulations for d+Au Running Trigger Board Meeting Dec5, 2002 MC & TU.
Advertisements

UK egamma meeting, Sept 22, 2005M. Wielers, RAL1 Status of Electron Triggers Rates/eff for different triggers Check on physics channels Crack region, comparison.
Tracey Berry1 Looking into e &  for high energy e/  Dr Tracey Berry Royal Holloway.
Current limits (95% C.L.): LEP direct searches m H > GeV Global fit to precision EW data (excludes direct search results) m H < 157 GeV Latest Tevatron.
Digital Filtering Performance in the ATLAS Level-1 Calorimeter Trigger David Hadley on behalf of the ATLAS Collaboration.
INTRODUCTION TO e/ ɣ IN ATLAS In order to acquire the full physics potential of the LHC, the ATLAS electromagnetic calorimeter must be able to identify.
Adding electronic noise and pedestals to the CALICE simulation LCWS 19 – 23 rd April Catherine Fry (working with D Bowerman) Imperial College London.
1 N. Davidson E/p single hadron energy scale check with minimum bias events Jet Note 8 Meeting 15 th May 2007.
RHESSI/GOES Xray Analysis using Multitemeprature plus Power law Spectra. J.McTiernan (SSL/UCB)
Simulation Work at Nevis Jovan Mitrevski Columbia University DØ Workshop July 10, 2002.
RHESSI/GOES Xray Analysis using Multitemeprature plus Power law Spectra. J.McTiernan (SSL/UCB) ABSTRACT: We present spectral fits for RHESSI and GOES solar.
1Calice-UK Cambridge 9/9/05D.R. Ward David Ward Compare Feb’05 DESY data with Geant4 and Geant3 Monte Carlos. Work in progress – no definitive conclusions.
1 N. Davidson E/p minimum bias update with Athena Analysis Meeting 12 th June 2007.
Saclay, 04 / 11 / 2002 E. Perez 1 Simulation of the performances of the upgraded L1Cal  Short reminder  Performances (jets) (see Jovan’s talk for electron.
17 May 2007LCWS analysis1 LCWS physics analysis work Paul Dauncey.
H  FTK update Catalin, Tony FTK meeting, 9/7/2006.
ALICE EMCal Physics and Functional Requirements Overview.
1 N. Davidson Calibration with low energy single pions Tau Working Group Meeting 23 rd July 2007.
General Trigger Philosophy The definition of ROI’s is what allows, by transferring a moderate amount of information, to concentrate on improvements in.
February 19th 2009AlbaNova Instrumentation Seminar1 Christian Bohm Instrumentation Physics, SU Upgrading the ATLAS detector Overview Motivation The current.
E. Devetak - LCWS t-tbar analysis at SiD Erik Devetak Oxford University LCWS /11/2008 Flavour tagging for ttbar Hadronic ttbar events ID.
Tests with JT0623 & JT0947 at Indiana University Nagoya PMT database test results for JT0623 at 3220V: This tube has somewhat higher than usual gain. 5×10.
Energy Flow and Jet Calibration Mark Hodgkinson Artemis Meeting 27 September 2007 Contains work by R.Duxfield,P.Hodgson, M.Hodgkinson,D.Tovey.
LHCC Review, CERN, 19/10/99Paul Bright-Thomas, for Alan Watson 1 LVL1 Calorimeter Algorithm Updates Changes since the TDR: Greater “integration” of e/
LVL1 Workshop, CERN, 16/07/99Alan Watson (by proxy) ATLAS Jet Trigger Algorithm Performance Requirements:  Good jet E T resolution Sharp threshold Minimise.
Alan Watson Atlas Trigger Workshop, Amsterdam, 18-22/10/10 1 Calibration of L1Calo L1Calo Calibration Overview L1Calo Calibration Overview Calibration.
26 Apr 2009Paul Dauncey1 Digital ECAL: Lecture 1 Paul Dauncey Imperial College London.
QCD and Top backgrounds in W+jets and Rjets Alessandro Tricoli (CERN) on behalf of W+jets and Rjets groups 3 rd May 2013 W+jets and Rjets EB Meeting.
A Study of Electron Identification Jim Branson UCSD with collaborators from FNAL, UCSB & UCSD.
CaloTopoCluster Based Energy Flow and the Local Hadron Calibration Mark Hodgkinson June 2009 Hadronic Calibration Workshop.
1ECFA/Vienna 16/11/05D.R. Ward David Ward Compare these test beam data with Geant4 and Geant3 Monte Carlos. CALICE has tested an (incomplete) prototype.
Photon reconstruction and matching Prokudin Mikhail.
26 Apr 2009Paul Dauncey1 Digital ECAL: Lecture 3 Paul Dauncey, Imperial College London.
Overview of the High-Level Trigger Electron and Photon Selection for the ATLAS Experiment at the LHC Ricardo Gonçalo, Royal Holloway University of London.
CALOR April Algorithms for the DØ Calorimeter Sophie Trincaz-Duvoid LPNHE – PARIS VI for the DØ collaboration  Calorimeter short description.
Γ +Jet Analysis for the CMS Pooja Gupta, Brajesh Choudhary, Sudeep Chatterji, Satyaki Bhattacharya & R.K. Shivpuri University of Delhi, India.
Study on search of a SM Higgs (120GeV) produced via VBF and decaying in two hadronic taus V.Cavasinni, F.Sarri, I.Vivarelli.
Monitoring of L1Calo EM Trigger Items: Overview & Midterm Results Hardeep Bansil University of Birmingham Birmingham ATLAS Weekly Meeting 11/11/2010.
ATLAS and the Trigger System The ATLAS (A Toroidal LHC ApparatuS) Experiment is one of the four major experiments operating at the Large Hadron Collider.
Search for High-Mass Resonances in e + e - Jia Liu Madelyne Greene, Lana Muniz, Jane Nachtman Goal for the summer Searching for new particle Z’ --- a massive.
Issues with cluster calibration + selection cuts for TrigEgamma note Hardeep Bansil University of Birmingham Birmingham ATLAS Weekly Meeting 12/08/2010.
Jet Studies at CDF Anwar Ahmad Bhatti The Rockefeller University CDF Collaboration DIS03 St. Petersburg Russia April 24,2003 Inclusive Jet Cross Section.
Hardeep Bansil (University of Birmingham) on behalf of L1Calo collaboration ATLAS UK Meeting, Royal Holloway January 2011 Argonne Birmingham Cambridge.
A search for the ZZ signal in the 3 lepton channel Azeddine Kasmi Robert Kehoe Southern Methodist University Thanks to: H. Ma, M. Aharrouche.
L1Calo EM Efficiencies Hardeep Bansil University of Birmingham L1Calo Joint Meeting, Stockholm 29/06/2011.
Trigger study on photon slice Yuan Li Feb 27 th, 2009 LPNHE ATLAS group meeting.
Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 In situ jet energy calibration General considerations The different.
10 January 2008Neil Collins - University of Birmingham 1 Tau Trigger Performance Neil Collins ATLAS UK Physics Meeting Thursday 10 th January 2008.
18 Sep 2008Paul Dauncey 1 DECAL: Motivation Hence, number of charged particles is an intrinsically better measure than the energy deposited Clearest with.
RHIC-PV, April 27, 2007 M. Rijssenbeek 1 The Measurement of W ’s at the CERN and FNAL hadron colliders W ’s at RHIC ! W ’s at CERN – UA2 W ’s at FNAL -
Mark OwenManchester Christmas Meeting Jan Search for h ->  with Muons at D  Mark Owen Manchester HEP Group Meeting January 2006 Outline: –Introduction.
ATLAS and the Trigger System The ATLAS (A Toroidal LHC ApparatuS) Experiment [1] is one of the four major experiments operating at the Large Hadron Collider.
Using direct photons for L1Calo monitoring + looking at data09 Hardeep Bansil University of Birmingham Birmingham ATLAS Weekly Meeting February 18, 2010.
L1Calo EM Efficiency Maps Hardeep Bansil University of Birmingham L1Calo Weekly Meeting 07/03/2011.
S. Dasu, University of Wisconsin February Calorimeter Trigger for Super LHC Electrons, Photons,  -jets, Jets, Missing E T Current Algorithms.
Study of missing Level-1 triggers using data10 Hardeep Bansil University of Birmingham Trigger E/Gamma Signature Group Meeting 20/05/2010.
Photon purity measurement on JF17 Di jet sample using Direct photon working Group ntuple Z.Liang (Academia Sinica,TaiWan) 6/24/20161.
E. Soldatov Tight photon efficiency study using FSR photons from Z  ll  decays E.Yu.Soldatov* *National Research Nuclear University “MEPhI”
Moriond 2001Jets at the TeVatron1 QCD: Approaching True Precision or, Latest Jet Results from the TeVatron Experimental Details SubJets and Event Quantities.
Introduction 08/11/2007 Higgs WG – Trigger meeting Ricardo Gonçalo, RHUL.
The LHCb Calorimeter Triggers LAL Orsay and INFN Bologna.
Phase2 Level-0 Calo Trigger ● Phase 2 Overview: L0 and L1 ● L0Calo Functionality ● Interfaces to calo RODs ● Interfaces to L0Topo Murrough Landon 27 June.
TrigEgamma Note L1 Plots Status Update Hardeep Bansil University of Birmingham Birmingham ATLAS Weekly Meeting 30/09/2010.
Performance of jets algorithms in ATLAS
L1Calo Requirements on the DPS
The University of Manchester
ATLAS L1Calo Phase2 Upgrade
Higgs → t+t- in Vector Boson Fusion
Plans for checking hadronic energy
Ivan Hollins 08/05/06 The University of Birmingham
Presentation transcript:

Alan Watson L1Calo Upgrade Meeting 1 EM Rejection in Phase1 Developments since Stockholm: Using depth information aloneUsing depth information alone Using transverse granularityUsing transverse granularity EM-Jet DisambiguationEM-Jet Disambiguation

Alan Watson L1Calo Upgrade Meeting 2 Methodology A brief MC study – not much changed since Stockholm  Form towers by summing CaloCells.  Keep finer-granularity subsums as well as complete tower sums  Enchanced transverse granularity plus depth information.  No cell noise cuts applied. No simulation of noise from layer summing.  1,500 MC10 W → e as signal, 1,700 JF17 as background  Medium electrons used for signal efficiency study  Pileup included (46 mbias/crossing) Algorithm simulation  Current EM trigger, with standard (analogue) inputs  Same algorithm on digital inputs  Finer-granularity sums formed at same time  Match “digital” RoI to “analogue” and combine features

Alan Watson L1Calo Upgrade Meeting 3 Transverse Granularity at L0? What might be possible?  FEB unchanged  4-channel sums in shapers  Told summed in  direction in mid layer – need confirmation in endcap  Change Layer Sum/Backplane as well as Tower Builder  Digital sums with transverse granularity as well as depth Simulated granularities  PS: 0.1×  /32  Strip/Mid: 0.025×  /32  Back: 0.05×  /16  Note: For the purpose of the study - not assuming all of these will be available.

Alan Watson L1Calo Upgrade Meeting 4 Fine-Granularity Algorithms RoI location based on current algorithm (2x2 core = max) Most energetic layer 2 cell (within central 2x2 region) Most energetic neighbour in phi (above or below) Add neighbours in eta to form cluster Wider eta environment for isolation/rejection Add overlapping cells in other layers to form E T cluster

Alan Watson L1Calo Upgrade Meeting 5 Jet Vetoes Studied Current L1Calo (analogue inputs)  EMIsol, HadIsol, HadCore – cuts on E T values Depth Only  EM back sample E T – cut on E T, fraction of EM cluster (digital)  Digitised at 100 MeV/layer. No negative layer E T. Transverse Granularity  Various shower width tests in layers 1 and 2  e.g. ratio of E T in 3x2/7x2 cells in layer 2  “L0 cells” digitised with 50 MeV count, no negative cell E T.

Alan Watson L1Calo Upgrade Meeting 6 Caveats Background dataset: JF17  QCD events that have been filtered at 4-vector level to exclude events highly unlikely to pass triggers  Possibility that these are biassed to narrower jets/denser cores. Rejection might differ slightly with minBias. Choice of cuts: signal statistics  Rejection is sensitive to precise cut value.  Statistical fluctuations in signal sample may lead to looser/tighter cut giving required efficiency.  Beware of making too fine distinctions from these data.

Alan Watson L1Calo Upgrade Meeting 7 Single Cuts. EM23 RoIs, Signal Efficiency ≈ 98% Cut variable and valueSignal εJF17 Survivial EM Isolation ≤ Had Isolation ≤ Had Core ≤ EM Cluster, Back Sample < 1.0 GeV EM Cluster, Back Sample/Total < EM Layer 2, 3×2/7×2 > EM Layer 1, 2x2/4x2 >  Typically ~5% statistical uncertainty on background rejection  Showing only cluster width definitions that give best performance in each layer  EM back layer fraction cut requires very fine tuning (sub- percent)

Alan Watson L1Calo Upgrade Meeting 8 Single Cuts. EM23 RoIs, Signal Efficiency ≈ 95% Cut variable and valueSignal εJF17 Survivial EM Isolation ≤ Had Isolation ≤ Had Core ≤ EM Cluster, Back Sample < 0.5 GeV EM Cluster, Back Sample/Total < EM Layer 2, 3×2/7×2 > EM Layer 1, 2x2/4x2 >  Note that had core cut cannot be tightened to 95% efficiency  EM layer 2 cluster width cut clearly most powerful now

Alan Watson L1Calo Upgrade Meeting 9 Two-Cut Combinations, Signal Efficiency ≈ 98% Cut variable and valueSignal εJF17 Survivial Had Core ≤ 1 + EM Isolation ≤ Had Core ≤ 1 + EM Back/Total < Had Core ≤ 2 + EM2 3×2/7×2 > Had Core ≤ 1 + EM2 3×2/7×2 > Had Core ≤ 1 + EM1 2×2/4×2 > EM2 3/7 > EM Back/Total <  Cluster width cuts show useful gains in rejection  But quite sensitive to cut value – arithmetical precision required

Alan Watson L1Calo Upgrade Meeting 10 Two-Cut Combinations, Signal Efficiency ≈ 95% Cut variable and valueSignal εJF17 Survivial Had Core ≤ 1 + EM Isolation ≤ Had Core ≤ 1 + EM Back/Total < Had Core ≤ 1 + EM2 3×2/7×2 > Had Core ≤ 1 + EM1 2×2/4×2 > EM2 3/7 > EM Back/Total <  Gains from cluster width more significant – mid layer width cut dominates rejection

Alan Watson L1Calo Upgrade Meeting 11 Three-Cut Combos, Signal Efficiency ≈ 98% Cut variable and valueSignal εJF17 Survivial Had Core≤1+EM Isolation≤ 3+Had Isol≤ Had Core≤1+EM Isolation≤10+EM2 3/7> Had Core≤1+EM Isolation≤ 4 +EM2 3/7> Had Core≤2+EM2 3/7 > EM1 2/4> HadCore≤2+EM2 3/7>0.89+EM Back/Tot< Had Core≤1+EM Isol≤ 4 +EM Back/Tot<  No real gain over the two cut combinations for same efficiency  Question simplicity vs robustness?  Best-performing combinations dominated by 2 cuts

Alan Watson L1Calo Upgrade Meeting 12 Three-Cut Combos, Signal Efficiency ≈ 95% Cut variable and valueSignal εJF17 Survivial Had Core≤0+EM Isolation≤ 2+Had Isol≤ Had Core≤2+EM Isolation≤ 5+EM2 3/7> Had Core≤1+EM Isolation≤ 4 +EM2 3/7> Had Core≤2+EM2 3/7 > EM1 2/4> HadCore≤2+EM2 3/7>0.89+EM Back/Tot< Had Core≤1+EM Isol≤ 4 +EM Back/Tot<  Again, little if any gain over two cut combinations.  Combinations including mid-layer width cut distinctly better than others

Alan Watson L1Calo Upgrade Meeting 13 Rate Comparison (unnormalised) – ε = 98% x2.5x3.5

Alan Watson L1Calo Upgrade Meeting 14 Rate Comparison (unnormalised) – ε = 95% x2.6 5 x4.5

Alan Watson L1Calo Upgrade Meeting 15 Comparison with Denis’/Steve’s Results Cuts for given efficiency slightly looser  Hence rejection is not quite as good. Possible reasons  Data preparation?  Calibration or noise handling differences  Cluster seeding?  My layer 2 cluster location is partly determined by L1 algorithm, rather than maximum being entirely determined by layer 2 cells  Datasets or statistics?

Alan Watson L1Calo Upgrade Meeting 16 Quick Cross-Checks Compare with L2 variable  Use T2CaRcore variable in ntuple  Match RoI word to L1 RoI  Results: 98% (95%) efficiency ⇒ 25% (24%) JF17 survival  Compared with 37% (28%) above  But still not quite as good as Denis saw – difference due to dataset, analysis? Sensitivity to cell noise cuts  Repeat with layer 2 4-cell sums truncated to 250 MeV counts  Results: 99% (96%) efficiency ⇒ 32% (23%) JF17 survival  Actually slightly better for coarser digitisation!  Cut values were slightly harder, presumably noise suppression effect  Would need to check effect for other RoI E T values.

Alan Watson L1Calo Upgrade Meeting 17 Algorithm Effects: Layer 2 Seeding Previous seeding was constrained by L1 algorithm  Find RoI location using current algorithm  Look for maximal cell within layer 2 inside 2x2 tower core region Remove this constraint on seeding  Just look for maxima within layer 2  Match RoIs found this way to L1 RoIs Very rushed  Last thing before holiday!  Very limited statistics (few hundred signal, ~1k background events)

Alan Watson L1Calo Upgrade Meeting 18 Effect of Purer Layer 2 Seeding Removing constraint does sharpen efficiency curve As used in studies above Pure layer 2 seeding  Also seems to produce slightly better rejection  Pretty similar to L2 algorithm. Very preliminary study though.

Alan Watson L1Calo Upgrade Meeting 19 Tentative Conclusions Concrete gains possible from ECAL transverse granularity  Not quite as strong as reported by Denis & Steve (S)  Algorithm differences seem to be partial explanation.  Combining 3/7 cell cluster fraction with hadronic isolation most powerful  Modest gains from adding third cut  Not tested systematically at lower RoI E T  Greater gain from tightening 98% → 95% efficiency that current L1 cuts  Signs of greater gains at lower RoI E T  Need confirmation, ideally with minBias sample (check for filter bias)?Caveats  Low-stats study, not tried to optimise tower noise cuts for lumi  Fine granularity implementation not fully realistic (RoI definition)  Precise (percent-level) precision used in fraction calculations

Alan Watson L1Calo Upgrade Meeting 20 EM-JET DISAMBIGUATION

Alan Watson L1Calo Upgrade Meeting 21 The Problem of Combined Triggers Current L1 uses only multiplicities  So if I want an EM + Jet trigger, or EM + TAU, how do I ensure these are not the same object? Easy if both have same E T  Any EM20 passes J20, so ask for EM20 + 2J20 and all is well Also OK if EM more energetic than jet  But that’s useless in practice! Tricky when jet more energetic than EM  Best you can do is something like EM20 + 2J20 + J50  …but even then, nothing stops the EM20 and J50 being same object  Isolation potentially complicates this (but I think issues overstated)

Alan Watson L1Calo Upgrade Meeting 22 Our Original Phase 1 Proposal Resolve ambiguities (in TP or CMX)  Match EM/TAU/Jet RoIs  Decide whether a distinct pair passes the trigger requirement  Determined that only modest coordinate precision (jet element size) needed. But what gain does it bring us?  Depends on trigger menu, of course  But have we ever actually studied this?

Alan Watson L1Calo Upgrade Meeting 23 Another Quick and Dirty Study Same JF17 samples as before (  ×filter =  b)  Choose some baseline threshold – 10 or 20 GeV  Normalise to events passing balanced combination trigger  EMx + 2Jx, EMx + 2TAUx, TAUx + 2Jx  Include isolated EM, TAU  Look at rate vs ET of more inclusive object (Jet or TAU)Disambiguation  Find most energetic TAU/Jet distinct from EMx/TAUx  Repeat for all EMx/TAUx RoIs, to find highest-ET disambiguated TAU/Jet in event  Plot fraction of events passing disambiguated trigger  Normalised to balanced combination trigger, as above  Estimate improvement in rate from disambiguation

Alan Watson L1Calo Upgrade Meeting 24 EM10+Jet vs Jet ET. No disambiguation 20 2E34

Alan Watson L1Calo Upgrade Meeting 25 Effect of Disambiguation – EM10 + Jet Main gain is when jet ET is 2-3xEM threshold At high ET most events have another jet passing EM10

Alan Watson L1Calo Upgrade Meeting 26 Effect of Disambiguation – EM10I + Jet Similar gains at mid ET Still merge at high ET

Alan Watson L1Calo Upgrade Meeting 27 Rate Improvement vs E T  Statistics poor, but indication that gains larger for more realistic EM20I trigger  Isolation also more effective EM10: gain bit under factor 2 at best EM20I: gain almost factor of 4. No statistics at higher ET

Alan Watson L1Calo Upgrade Meeting 28 EM+TAU Disambiguation Harder problem, as objects more similar Gain ~20% over broad range of TAU ET

Alan Watson L1Calo Upgrade Meeting 29 TAU + Jet More like EM + Jet

Alan Watson L1Calo Upgrade Meeting 30 Fine-Grain Isolation Plus Disambiguation The rejection from isolation alone seems large…

Alan Watson L1Calo Upgrade Meeting 31 Fine-Grain Isolation Plus Disambiguation Fractional gain better than with weaker isolation

Alan Watson L1Calo Upgrade Meeting 32 More Tentative Conclusions EM/Tau-Jet Disambiguation  Could be useful, even promising, if kinematic range between EM/Tau and jet not too large  Hints that stronger isolation (better jet rejection) improves this EM-Tau Disambiguation  More difficult to make major gains over current solution.  Could still be useful in making efficiencies more comprehensible Need example use cases  And more statistics!