A vertex trigger for LHCb Vertex 2002 Hawaii
A vertex trigger for LHCb Vertex 2002 Hawai’i, 7 Nov 2002 Niels Tuning (CERN) (on behalf of LHCb) The trigger for LHCb …. . and the use of the Si vertex detector at the first and second trigger levels 7 Nov 2002 Niels Tuning - Vertex 2002 1
LHCb A Large Hadron Collider Beauty Experiment for Precision Measurements of CP-Violation and Rare Decays Colliding beams: n n n n 25 ns 7 Te. V x 7 Te. V pp L = 2. 1032 cm-2 s-1 (visible) = 68 mb (pp b b. X) = 0. 5 mb ~1012 b b / year BR(interesting channels) ~10 -2 – 10 -9 Finding B-mesons: n n n 7 Nov 2002 High PT decay products Large lifetime sec. vertex Invariant mass Niels Tuning - Vertex 2002 A low multiplicity B + - event LHCb trigger = looking for a needle in a haystack… …every 25 ns! 2
LHCb Trigger – Overview L 0: high PT n Pile-up Veto, using VETO detector n n High ET calorimeter objects High PT muons L 1: high PT + impact parameter n High impact parameter tracks, using VELO detector n High PT tracks, using TT detector and L 0 info Bunch crossing rate 40 Mhz Non empty bc rate 30 MHz Visible interaction rate ~12 MHz Input to L 0 ~10 MHz Input to L 1 1 MHz Input to higher level 40 k. Hz Writing to disk 0. 2 k. Hz L 2+L 3: high PT + displaced vertex + B-mass + PID n Use tracking stations and RICH 7 Nov 2002 Niels Tuning - Vertex 2002 3
TT LHCb detector T 1 T 2 T 3 L 0 veto 7 Nov 2002 L 1 L 0 trigger Niels Tuning - Vertex 2002 4
L 0 – Pile-up VETO (L 0 = first trigger level) 320 strips Purpose: remove multiple interactions n n Nominal luminosity: L = 2. 1032 cm-2 s-1 Single : Double : Triple 16 : 4 : 1 75% : 20% : 5% n More difficult to find high IP tracks at L 1 Reduce bandwidth for L 0 Detector: n n n 2 Si disks (4 sensors) Same sensors as VErtex LOcator (see talk J. Palacios) Only R information Use Beetle chip w OR of 4 strips: comparator output of 4 channels 1280 channels for 2 disks 7 Nov 2002 Niels Tuning - Vertex 2002 84 mm n 16 mm Why? 5
Rb (cm) L 0 – Pile-up VETO algorithm true all Calculate vertex for all combinations of 2 points a and b. Find highest peak (PV) Remove the hits and find 2 nd peak Veto if peak>threshold (Zvtx) 2. 8 mm, (beam) 53 mm combinations Ra (cm) Z vtx (cm) 7 Nov 2002 Niels Tuning - Vertex 2002 6
L 0 – Pile-up VETO performance B + - L 0 efficiency increase n n from 50% to 60% the L 0 PT, HADR threshold can be lowered from 4 Ge. V to 3. 6 Ge. V Same L 0 output rate! Reduce bandwidth and enhance purity: n n Pileup VETO vetoes ~15% of all events Vetoed events are more likely to trigger Only small inefficiency for single interactions: ~5% Reject ~30% of multiple interactions (NB: multiple interactions include inelastic+elastic !) 7 Nov 2002 Niels Tuning - Vertex 2002 7
L 1 trigger: vertex trigger (L 1 = second trigger level) Implementation: n Clustering in FPGA on front-end CPU w Send data to RU (3 -4 GB/s) n CPU-farm: w 300 – 400 CPUs w 2 D torus n n Use scheduler Prototype with 32 CPUs running at 1. 24 MHz SCI RU Scheduling network Buffer depth: 1820 events Latency = 1. 65 ms Strategy: n n n Find 2 d-tracks with R-sensors and reconstruct vertex Reconstruct high-impact parameter tracks in 3 d Extrapolate to TT through small magnetic field PT Match track to L 0 Muon objects PT and PID Select B–events using impact parameter and PT information 7 Nov 2002 Niels Tuning - Vertex 2002 8
L 1: VErtex LOcator (see talk by Juan Palacios) Si: 220 m thick, n-on-n, n Pitch: 37– 98 m, R 40– 92 m n sensor 2048 strips R sensor 2048 strips Sens. area: 0. 8 cm < R < 4. 2 cm 21 stations (84 sensors) n n -17. 5 cm < Z < 75 cm 170, 000 channels RF foil: Very thin “beampipe” to separate prim. and sec. vacua sensor: stereo angle +10 o, -20 o 7 Nov 2002 Niels Tuning - Vertex 2002 R sensor: 4 inner and 2 outer sectors 9
L 1: input data Cluster resolution: (testbeam) Velo clusters: n n Clusters are found in FPGAs per groups of 32 strips. Digital (offline is analog) ~1000 clusters ~0. 1% noise clusters ( 200) Resolution ( m) =14 m Pitch ( m) TT clusters n ~300 clusters 1000 clusters L 0 objects n n 7 Nov 2002 (simulation) 3 muons Some calorimeter data Niels Tuning - Vertex 2002 10
L 1: track reconstruction Look only for R-clusters: 2 d RZ-tracks n Fast! Find “triplets” of clusters Combine triplets ~98% efficiency for B-tracks ~60 multiplicity 2 d tracks in a 90 o sector: Z vtx histogram 7 Nov 2002 X, Y vtx Niels Tuning - Vertex 2002 VELO is being redesigned to 45 o sectors: n faster L 1 tracking n lower noise n less 2 d tracks 11
L 1: primary vertex Primary vertex reconstructed with 2 d tracks XY information from segmentation Flight direction of B is forward RZ projection of impact parameter contains most information Impact parameter: 2 d Vertex resolution: 3 d X Z Lifetime 7 Nov 2002 Niels Tuning - Vertex 2002 12
L 1: PT information – L 0 CAL Match VELO track to Muon from L 0: n PID n Momentum VELO TT MAGNET Efficient selection of n Bs J/ ( ) n Bd J/ ( ) Ks dp/p=4. 8% Enhance tagged sample 7 Nov 2002 MUON Niels Tuning - Vertex 2002 13
L 1: algorithm (1) Preliminary Good momentum resolution, cut on J/ mass: n n J/ mass ~60% of events contain both muons <1% min. bias retention OR require 1 muon with high PT, high IP n n ~75 % efficiency ~3% retention Performance with all event info (except TT) Work ongoing: n 7 Nov 2002 achieved 90% eff. using net neural Niels Tuning - Vertex 2002 14
L 1: Trigger Tracker VELO Design still under study Si: 400 - 500 m thick Wide pitch: 200 m TT TT = Tracking station before the magnet MAGNET RICH-1 141 cm Sensor dimensions: 7. 8 x 11. 0 cm 2 n n Stereo angle: 00, -50, +50, 00 To be optimized 836 sensors (~7 m 2) 7 Nov 2002 Niels Tuning - Vertex 2002 116 cm 4 layers (x, u [30 cm gap] v, x) 15
L 1: PT information – TT High impact parameter 2 d tracks are reconstructed in 3 d and extrapolated to TT 1 B dl 0. 1 Tm Ensures momentum information n 30% --- 30 Ge. V --- n dp/p Magnetic field between VELO and TT: dp/p ~ 30% 7 Nov 2002 Niels Tuning - Vertex 2002 p (Ge. V) 16
L 1: algorithm (2) Preliminary Get two highest PT tracks, using TT Consider impact parameter and PT of these tracks Look in plane IP/ (IP) vs PT Bd + - Bs Ds-K+ Signal Minimum bias 7 Nov 2002 Niels Tuning - Vertex 2002 17
n n Possibly x 32 more. A more flexible system is under study were CPUs from DAQ can be used for the TRIGGER and vice versa Tracking+Vertexing: < 20 ms 2007 CPUs: x 8 faster Optimize algorithm+code in the right ballpark! 2 d tracking: ~70% Vertex: ~15% 3 d tracking (a few) ~15% Tracking+Vertexing ---- 17 ms Remember: latency ~1. 7 ms Events L 1: performance - timing time (ms) 7 Nov 2002 Niels Tuning - Vertex 2002 18
L 1: performance L 1 output rate (MHz) Efficiency vs retention: (Example B + - ) Bd + - Expected overall trigger performance: (cumulative) L 1+L 0 info 40 k. Hz L 1+PT info Bd + - efficiency 7 Nov 2002 Niels Tuning - Vertex 2002 L 0 L 1 L 2 19
Conclusions The pileup-VETO detector efficiently rejects multiple primary vertices @ 40 MHz at L 0 The VELO detector reconstructs primary vertices at L 1 with excellent resolution n n (Zvtx) 60 m high impact parameter tracks can be identified The TT detector - or L 0 information - enables measuring the momentum of tracks Efficient L 1 selection algorithms under study n Efficiency of 70% (90%) for B + - ( B J/ ( ) Ks) reachable at 4% minimum bias retention 7 Nov 2002 Niels Tuning - Vertex 2002 20
- Slides: 20