Four-Vector Track-Trigger Studies

1. Four-Vector Track-Trigger Studies Jim Brooke, Emyr Clement, Giulia Ferlito, Robert Frazier, Kristian Harder, Lucy Kogan, Dave Newbold 23/07/09 Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

2. Overview of 4-Vector Track-Trigger Studies Tool • History + Effort: • Originally written by Jim Brooke; now also Kristian Harder and myself. • Additional effort in the form of 3 summer students for the next ~2 months. • Purpose: • Fast and easy-to-use tool to scan over geometry and trigger options/ideas • If it doesn’t work at the 4-vector level, it will never work. • Else… move onto fast/full simulation. • What it does: • Extrapolates Pythia-generated particles on exact helices. • Simulates detector geometry: cylindrical layers with modules and pixels. • A hit is generated at the centre of each pixel traversed by a track helix. • Proper treatment of looping tracks. • Approximates secondary particle production through material effects. Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

3. Overview of 4-Vector Track-Trigger Studies Tool • Advantages: • Very fast: • 0.003 seconds/event no pile-up. • ~ 5 seconds/event with 200 pile-up. • Detector geometry can be changed in seconds. • Also supports short layers. • Code is fairly self-contained, no major headaches in adding new features. • Not yet implemented: • Clustering • Tracks from stubs (in progress - should have v1 soon) • Endcaps (also in progress) Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

4. Mission • To untangle the purely physicaleffects which limit the performance from the detectoreffects. • Understand the irreducible performance limit. • Provide a cross-check for full-/fast-sim studies. • Define limits of performance these studies can reach when fully optimised. • Provide a rapid indication of the effects of changes that are hard to validate in more detailed simulation code. Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

5. In Detail - Geometry • Single, simple geometry file that is easy to change: • Barrel layers, modules and pixel dimensions. • E.g. The Long Barrel geometry (seen by hits) • No short stacks: r/cm z/cm Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

6. In Detail - Geometry • Single, simple geometry file that is easy to change: • Barrel layers, modules and pixel dimensions. • E.g. The Long Barrel geometry (seen by hits) • No short stacks: • With short stacks: r/cm z/cm Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

7. In Detail - Geometry • Single, simple geometry file that is easy to change: • Barrel layers, modules and pixel dimensions. • E.g. The Long Barrel geometry (seen by hits) • No short stacks: • With short stacks: • Adding endcaps into the simulation is now a priority. • Essential in order to study the hybrid geometry. r/cm z/cm Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

8. r/cm z/cm In Detail - Track Propagation • Now uses the FastSim propagator class • BaseParticlePropagator • Looping tracks fully supported: • r-z view, 0.7 GeV particle gun tracks looping down the length of the barrel: Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

9. In Detail - Material Interactions • Simulation of secondaries: • Currently, secondary particles are not linked directly to tracks crossing a layer. • Estimate number of secondary interactions as: • Create corresponding number of interaction vertices with random distribution across layer. • Create two charged particles for each production vertex. • Still need to tune the momentum distribution of the secondaries. Thickness of layer in radiation lengths (configurable)  Interactions per radiation length (currently set to 1)  Estimate of number of charged tracks crossing a layer Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

10. In Detail - Material Interactions • Distribution of MC particle vertices (radius in x-y plane) Material effects of long stacks 1,2,3,4,9,10 clearly visible r/cm Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

11. In Detail - Stub Finding • Stubs are found on a per-module basis • Local algorithm. • Inner-hit pixel on a stack is the centre of a configurable phi/z acceptance window for the outer-hit. • Acceptance window specified in number of pixels in phi and z. • E.g. for a 1 stub acceptance window: Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

12. In Detail - Stub Finding • Local stub-finding algorithm produces a stub-threshold that varies with stack radius. • Threshold moves from ~0.5 GeV to 2 GeV for a phi acceptance window of 2 pixels. (Long barrel, 1mm stack separation). Pt (GeV) Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

13. In Detail - Stub Finding • Local stub-finding algorithm produces a stub-threshold that varies with stack radius. • Threshold moves from ~0.5 GeV to 2 GeV for a phi acceptance window of 2 pixels. (Long barrel, 1mm stack separation). • Slightly problematic for FastSim comparisons • Fast-sim stub-finder currently seems rather artificial. • Uses floating point phi cut for stub-thresholds? Pt (GeV) Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

14. Comparisons with Fast Simulation • Apples to apples comparison with results from Laura Fields’s work using the Fast Simulation • Good agreement between the two simulations for hit- and stub-rates • Using: • Long barrel geometry (without short stacks). • 1mm stack spacing. • Stub-thresholds are matched as well as possible. • Material interactions not switched on in four-vector simulation (needs more tuning). • 200 pile-up events per crossing, no signal. Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

15. Comparisons with Fast Simulation • Sample hit-rate comparisons: Four-Vector Fast-Sim Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

16. Comparisons with Fast Simulation • Sample stub-rate comparisons: Four-Vector Fast-Sim Request: FastSim results with better statistics to compare against! Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

17. Results Wiki Page • We now have a wiki page for results: • https://twiki.cern.ch/twiki/bin/view/CMS/FourVectorSLHCTriggerSimulationResults • Similar to Laura Fields’s “standard plots” results page: • http://www.lns.cornell.edu/~ljf26/SLHCTrackingTriggerStudies/standardPlots.html • Currently only shows results for variations of the long-barrel geometry. • Variations in stack separation and stub-finder pixel window in phi. Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

18. Future Plans • We now have a lot more effort available • Myself up to ~80% on this now. • Lots of summer student effort available. • In good standing for getting results by September. • Critical software tasks to finish: • Add in endcaps ASAP. • Add algorithms to generate tracks from stubs. • Link to FastSim Calo + Muons. Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

19. Future Plans • Once all the software is in place… • We can quickly cover large amounts of geometry phase-space. • Study track-matching algorithms for different geometry options: • Particularly outside-in approach (not well covered yet?) • Optimise efficiencies • Examine response of track-matching algorithms for fake electrons located by calo trigger. • Can we reject this reliably? • Understand tradeoff between rate-reduction and efficiency for the parameters of our detector and algorithms. • Quantify occupancies and data rates. Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies

20. For Late September • Full scan over parameter space • Long Barrel. • Hybrid Geometry. • Others? • Performance with realistic quantities of readout data per primitive • Estimates of trigger rate reduction • For a range of stub-matching algorithms Robert Frazier -- University of Bristol -- 4-Vector Track-Trigger Studies