1. The University of Manchester
Triggering Top Events
Simon Head
The University of Manchester
Presenting on behalf of:
Thorsten Wengler (Manchester),
Attila Krasznahorkay (Debrecen / CERN),
David Berge (CERN),
Juergen Thomas (Birmingham),
Alan Watson (Birmingham)

2. Motivation for Top Physics at ATLAS
ATLAS will be a top factory
Cross-section for top production at ATLAS is ~800 pb
In 10 fb-1 (one years low luminosity running) at 14 TeV we expect 8 million tops
Large mass of the top quark makes it unique and sensitive to new physics

3. Triggering
Collisions at the LHC occur at a rate of up to 40 MHz, compared to a storage capacity of about 200 Hz.
The trigger system has the “interesting” task of rejecting % of the collision events while keeping those needed to achieve the physics goals of ATLAS.
Internally the trigger identifies and makes use of Regions-of-Interest
Each level of the trigger refines the decision made by the previous level and has access to more detailed information about the event
ATLAS uses a 3 level trigger system
First Level Trigger (LVL1)
High Level Trigger (HLT) - consisting of:
Second Level Trigger (LVL2) and
Event Filter (EF)

4. The Atlas Trigger System
The LHC proton bunches will cross at a frequency of 40 MHz
At the design luminosity of 1 × 1034 cm−2 s−1, roughly 23 inelastic proton–proton collisions at each bunch crossing
The LVL1 trigger will reduce the initial event rate to less than 75 kHz
The HLT must reduce the event rate further, to O(100) Hz

5. Trigger and Top
The ATLAS trigger is designed to fire on high pT leptons, jets and missing ET
The semi-leptonic ttbar decay provides all of these
Athena is the first release suitable for detailed trigger studies of fully simulated data
Data samples are produced for the CSC (Computing System Commissioning)
The ATLAS collaboration will write a number of CSC notes (produced by the Physics and Joint Performance group)

6. LHC nominal running b t / tbar l v q tbar /t qbar b
Concentrate on the semi-leptonic ttbar decays
Distinctive signal to trigger on (high PT lepton with isolation and missing ET)
Centrally produced (CSC) sample (14 TeV) – using “recotrig” AOD
Make use of TopView and EventView for event selection
Write custom tools to insert trigger data into the EventView

7. Environment for the study
Work presented here is in the context of note T5: “Triggering Top Quark Events in ATLAS”
Performance study
Exploit the trigger setup for efficient ttbar event selection
Define and test methodology to characterise trigger performance
e.g. Determine trigger efficiencies from data using simultaneously fulfilled trigger signatures in ttbar events.
Data must be officially produced with Athena or Re-run with 12.0.x using a misaligned geometry.
More detailed studies will be performed in the next two to three months using TopView

8. TopView Extensions to EventView to do top analysis
Reconstructs semi-leptonic decays for ttbar pairs by simply combining reconstructed W with a tagged b jet
We run with a modified version of the ttbar analysis included with TopView
We have written tools (EventView) that so far add:
LVL1 trigger information to the EventView and save to an NTUPLE
Full muon slice
Write out composite particles to the user data
Perform some simple analysis

9. Available Trigger information (11.0.5)
LVL1: The RoIs produced by the LVL1 system are available through the LVL1_ROI container. Information provided: eta and phi coordinates and passed threshold level. (Also: name of the threshold, and from the passed threshold value in GeV). Other quantities such as the CTP Decision are also available
LVL2: Among the provided information: eta and phi coordinates, and pT of the reconstructed particle
EF: Not available in

10. LVL1 egamma efficiency
What is the efficiency that a true electron will fire the First Level Trigger, I.e. pass a trigger threshold such as EM25i ?
EM25i asks for a single isolated electron with PT > 25 GeV
Common trigger object fired for many processes
Electrons within |eta| < 2.5
EM25i
EM cluster > 19 GeV
EM isolation <= 3 GeV
Hadronic core <= 2 GeV
Hadronic isolation <= 2 GeV
Start by plotting some simple quantities…

11. LVL1 egamma efficiency
PT spectrum for the electron in ttbar semi-leptonic decay (from truth)

12. LVL1 egamma efficiency EM25i as defined above
Threshold has reached about 90% efficiency at its nominal value of 25 GeV, as it should.

13. Electron Trigger in Top Events
Use teb to set isolation cuts
More activity than Zee
lower efficiency for same cuts
Isolation cuts for initial running:
EM isolation  3 GeV
Inner Had isolation  2 GeV
Outer Had isolation  2 GeV
But rates for 1031 very low
Tau trigger, ET > 9 GeV, no isolation  ~4 kHz
9 GeV unisolated electron would be a subset of this
Use very loose trigger initially to understand performance

14. Top Efficiency from Electron Trigger
Run 5200
At least 1 top  lepton decay
Unisolated EM trigger  high efficiency
including events with no electrons!
many rejected when isolation applied
Events containing > 25 GeV e
Typical Offline top selection
Used MC truth – no reconstruction inefficiency
Trigger efficiency very high, with or without isolation

15. Top Efficiency from Jet Trigger
All numbers very preliminary
Run 5200
Expect 4 low-mid ET jets/event
electrons can also fire jet trigger
1033 trigger menu not efficient
1 Jet > 200 GeV ~15%
3 Jets > 75 GeV  ~40%
see box for more explanation
But 1031 is a different story
1 Jet > ~100 GeV 80-90% efficiency, 200 Hz rate
Efficient redundant trigger possible
ttbar 6 jets
Fast simulation study suggests 5 jet trigger promising, even at 1033
can we use these offline though?
Plot shows actual L1 Jet ET
Scale does not match offline
Trigger threshold << offline ET
200 GeV jet  ~150 GeV threshold
75 GeV jet  ~ 40 GeV threshold
Usually quote Jet ET for which trigger is 95% efficient, rather than (calibration-dependent) threshold

16. Jet Efficiency vs. ET
Comparison of trigger performance against reconstructed cone 0.4 jets
efficiency for individual jets

17. Missing ET in Top Events
Inclusive ETmiss trigger?
Possible at 1031?
ETmiss > 40 GeV  60 Hz
excluding machine backgrounds
Gives ~60% efficiency for SL top events
Probably more interesting to use top events to understand ETmiss trigger
Maybe multijet + ETmiss could be used as a redundant control for muon triggers in top events?
ETmiss is a challenging trigger
Global variable, so hard for L2 to improve
Combine ETmiss with other trigger or require low L1 rate
Most vulnerable to machine backgrounds

18. Muons Work continued from last Trigger and Physics week
Working on getting larger statistics
Our analysis code has access to the full muon slice
The plots opposite are just a small selection

19. Level-1 Trigger Item Overlaps in 11.0.5 RecoTrig AODs
Goal of item overlap study: determine efficiencies of each item using simultaneously fulfilled trigger signatures
Level-1 Trigger menu in AODs:
(0) Electromagnetic: (EM) 15 GeV, no isolation
(1) Hadronic (HA) 20 GeV, no isolation
(2)-(7): MU of 6, 8, 10, 11, 20 and 40 GeV
(8): Jet (JE) 40 GeV, size 4x4
Practically all events are triggered ! Only 85 out of Ev are not. Producing ROIs for each trigger type.
Histo (top) shows trigger bits sets: Each of EM, HA and JE trigger nearly every event. Only MU shows pT dependence.
Legoplot (bottom) illustrates overlap of items (logical AND of bits)

20. Level-1 Trigger Item Overlaps: Trigger Menu 1033
Now set a trigger menu suitable for 10^33 (TDR-like items), Re-done Digi to AOD step (200 Ev). To study combined triggers.
Items:
Jet: (0) Single Jet 180 GeV, (1) Three Jets, 75 GeV each (2) Four Jets, 60 GeV each
Hadronic (tau): (3) 20 GeV, isolation 5/10 GeV,
Electromagnetic: (4) 15 GeV, isolation 3/2/2 GeV
Combined: (5) Hadronic 30 GeV + Missing-ET 30 GeV, (6) Jet 75 GeV + Missing-ET 30 GeV
Legoplot (bottom) illustrates overlap of items: Logical AND of item bitmap (Word0 of CTP Decision)
As expected, items (3) (5) (6) overlap strongly

21. Conclusions Now is a good time to do trigger aware analysis
The Trigger is extremely important and thus far, mostly overlooked in analysis
Information is in the AOD, especially those titled “recotrig”
We are developing the tools to allow access to the information for EventView analysis
Information is already there, if you know how to access it
These tools should make accessing it using EventView easier
For a luminosity of 1031 we can use an EM trigger without isolation
Some RoI’s seem to match an electron (energy) but are a long way from the electron in eta-phi – why?