Tracking at the ATLAS LVL2 Trigger

1
Tracking at the ATLAS LVL2 Trigger

• Nikos Konstantinidis
• University College London

Athens HEP2003
2
Outline

• Introduction
• ATLAS Trigger Strategy
• Tracking at LVL2
• The IDScan tracking package
• The algorithms
• Performance
• Conclusions Outlook

3
Triggering at the LHC

• Challenge 1
• Bunch crossing every 25ns gt rate 40MHz
• Data storage capability 100Hz
• Must select online a couple events in a
  million!!!
• Online background rejection of 6 orders of
  magnitude!
• Challenge 2
• Peak luminosity 2x1033(low) 1034
  (high)
• 5 25 pp interactions per bunch
  crossing
• Luminosity falls by a factor 2 over a fill
  (10hours)
• Interesting (high pT) pp interaction
  complicated by pile-up
• Very annoying for tracking (increases
  combinatorics)

4
ATLAS Trigger Strategy
5
ATLAS Trigger Overview

• LVL1
• Uses Calorimeters Muon Trigger Stations (coarse
  granularity)
• LVL2
• Uses LVL1 Regions of Interest (RoI), so only a
  small fraction of the event data is accessed
• InDet tracking avail.
• Combines sub-dets
• Full granularity
• Event filter
• Refined, offline-type reconstruction, with access
  to calibration alignment data

6
Region of Interest
x-y view
r-z view
7
Tracking _at_ LVL2

• Tracking is needed for
• Single, high-pT electron/muon identification
• Match tracks to info from outer detectors
• B Physics (at low lumi? budget permitting?)
• Exclusive reconstruction of golden decays (e.g.
  Bgtpp)
• b-jet tagging (e.g. in MSSM H gthh gt bbbb)
• All must be done in 10ms
• Must deal with combinatorics
• At high luminosity 20K space points in the Si
  Trackers

8
The Si Trackers of ATLAS
9
The IDScan algorithms

• A sequence of four algorithms for pattern
  recognition track reconstruction using 3D space
  points.
• Basic idea
• Find z-position of the interesting (high-pT) pp
  interaction before any track reconstruction
• Select only groups of space points consistent
  with the above z

10
ZFinder

• Relies on
• Tracks are straight lines in rz. Using the (r,z)
  from a pair of space points of a track, you can
  determine its z0 by simple linear extrapolation
• High-pT tracks are almost straight lines in rf.
• Steps
• Make very thin slices in f (0.2-0.3 degrees)
• In each slice, make all pairs of space points
  from different layers, calculate z0 by linear
  extrapolation and fill a 1D histogram with this
• The bin with the max. number of entries
  corresponds to the z0 you are looking for

11
ZFinder Example
Jet RoI from WH(120GeV)
12
HitFilter

• Given this z0, all space points of a track
  originating from z0 will have the same h
• Steps
• Put all space points in a 2D histogram in (h,f)
• Accept all space points in a bin if this bin
  contains space points in at least 5 (out of 7)
  different layers
• Reject all other space points

• No combinatorics gt linear time behaviour
• Returns groups of hits

13
HitFilter Example
x-y view
r-z view
h-f histogram
14
Performance (I)

• Single e (pT40GeV) RoI at high L (DhxDf
  0.2x0.2)
• lt space pointsgt 250
• ltexecution timegt 1ms
• ZFinder resolution 180mm
• Efficiency 98
• B physics (low L) full Si Trackers reconstruction
• ltexecution timegt 20ms

projected to CPU speed of 4GHz
15
Performance (II)

• Using 2 rather than 3 pixel barrel layers
• Only necessary to change the min. number of space
  points required to make a track, from 5 (out of
  7) to 4(out of 6)

Algorithms conceptually simple gt Flexible gt
Robust
16
Example Electron RoI





17
IDScan Virtues

• Modular gt flexible gt robust
• Fast and linear t ( of space points)
• Suitable for all tracking needs of LVL2
• No DetDescr dependence only space points
• Uniform treatment of barrel/endcaps
• Uniform treatment of pixel/SCT

18
Summary Outlook

• Triggering has a central role at the LHC the
  physics reach of ATLAS (and CMS) depends on it
  critically
• Tracking at LVL2 is a real challenge, especially
  at high luminosity
• Determining the z-position of the interesting pp
  interaction prior to any track reconstruction and
  then rejecting all space points that cannot be
  due to tracks from that z seems to work best
• Still space for novel ideas to improve the ATLAS
  physics potential and exploit the physics at the
  LHC optimally