2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 1

The LHCb trigger strategy and performance

OverviewLevel-0 Detectors, algorithms, performanceLevel-1 Detectors, algorithms, performanceOutlook

Note: Higher Level triggers Offline selection(not covered in this talk)

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 2

Overview of LHCb trigger

L0:fully synchronous custom-made electronics fixed latency: 4 sreduce visible rate by ~ 1:10

L1 & HLT:flexible implementation (all software)share the same farm ~1000 CPUsasynchronousaverage L1 latency ~ 0.5 ms

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 3

PYTHIA VISIBLEtotal 100 60 mbbeauty 0.6 0.2 mb

To tape: 3106 Bd evts/yr 1106 Bs evts/yr

Clock (40 MHz)

Bd +-

Eff

icie

ncy

reco

nstr

ucta

ble

even

tsE

vent

rat

es [

Hz]

Bs J/ (+-) (K+K- )

Bs Ds K

Trigger rates: overview

Effective reduction by L0

bunch crossing (30 MHz)with visible collisions (10 MHz)

not PU-vetoed (9 MHz)

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 4

L0 & L1: General idea

Looking for Bs DsK , J/ψ , Kπ , …Bd π π , π , J/ψ KS , Κ*γ , …

UseB-meson signatures:

B mass or poor man’s B mass ~ high PT

B lifetime ~ displaced verticesglobal variables to select cleanest B events and reduce fortuitous triggers:

Veto pile-up eventsCut on hit multiplicity

+ some channel-specific tricks:e.g. for dimuon (already at L0)

(rec. in L1)

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 5

Electromagnetic Calorimeter:Shashlik type, 5952 cells, each 8-bit ET, 25X0

Hadron Calorimeter: Fe & scint. tiles, 1468 cells, each 8-bit ET cells 5.6 I

Scintillating Pad Detector and Preshower: (SPD/PS)both 5952 cells (each 1 bit)

Pile-up Veto:4 Si sensors

Muon chambers:MWPCs, 26k pads, projective in y

Detectors in L0y

z

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 6

Pile-Up Veto: DetectorPurpose: give more bandwidth to pp crossings with single collision (good for B physics!)

Use same sensors as in Vertex Locator4 half-discs upstream of the interaction pointMeasure radius r (strip pitch 40 … 1XX m)PU veto

Hybrid

Beetles

Silicon

Each disc: 2048 real strips ORed by groups of 4 in Beetle FE chip (discriminator mode, 80 MHz LVDS binary output)Covers –4.2 < < -2.9

Beetles

DiodesRoutinglines

84 mm

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 7

Pile-Up Veto: principle

RB (c

m)

ZPV (cm)

RA (cm) RA (cm)

RB (c

m)

peak P1 (peak size S1)

peak P2 (peak size S2)

ZB ZA

RB

RA

ZPV

Silicon r-sensorsB A

allows locating and counting visible pp collisions

RA ZPV - ZA

RB ZPV - ZB=k

ZA - k ZB

1 - k or ZPV =

Tracks from same ZPV have the same ratio k

k

ZPV’

k’

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 8

Pile-Up Veto: Performance

Rate of single-collision B-eventsas a function of luminosity

Before level-0

After level-0: without and with Pile-up veto gain factor of at least 1.4 in B-event rate at the same nominal luminosity (2 1032)

cm-2 s-1

Size of 2nd Peak S2

200

100

0

200

100

0

Rat

e (a

.u.)

Rat

e (a

.u.)

Bd +-

Min. bias

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 9

L0 Decision & Bandwidth

DivisionL0 Decision Unit collects:

From calorimeters: ET of all candidates (hadron, electron, , etc.)ET (also of previous and next two crossings)SPD hit multiplicity

From muon trigger: 42 largest PT

From pile-up detector: number of tracks and z of 1st and 2nd vtxtotal hit multiplicity

DU performs simple arithmetic, withadjustable thresholds, downscaling, etc.Bandwidth Division:

L0 output rate fixed to 1 MHz (keep 100 kHz contingency).For every given off-line selected channel, scan over L0 thresholds and PU-Veto parameters to give best performance for this given channel. Then, minimise “combined” loss for all channels.

Bandwidth Division: example

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 10

Detectors in L1: Vertex Locator

-sensors

r-sensors

Purpose: find displaced verticesVELO: 220 m thick Si, 170k binary channels for L1.Each station is a sandwich of an r- and -sensor.

45o sectorsSensitive area starts at ~8 mm from beam axis.VELO mounted on XY-table to center on beams.Roughly 1000 clusters per event.

10o-20o stereo angle

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 11

L1 vertexingUse r-z tracks (~70/evt) to get primary vtxTracks with 0.2 < d/mm < 3 (~8/evt) are reconstructed in 3D using -sensors (d = impact parameter to primvtx (2D or 3D ?)

91 cm91 cm

3.4

cm3.

4 cm

x,yrms = 30 m

zrms = 75 m

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 12

ee

hh

L1: matching L0 objects

Re-use L0 candidates in L1Combine E/HCAL energy+cell position and Muon tracks with VELO tracks3D-reconstruct those with best match (~ 8/evt)??

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 13

Example: L0- match

wdce

If m > 2 GeV then L1-yes

( keep the rare decays!)

Matched L0-: p/p ~ 5%

p/p

3D tr

acks

Adds ~ 5 kHz L1 rate

Bs J/ (+-)

Grabs about 60% of the L0-yes

“Minimum bias”

(offline selected)

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 14

Detectors in L1: Trigger Tracker

Purpose: rough momentum measurement (p/p~0.2) for L1.

0 o –5 o 5 o 0 o

Stereo

4 silicon planes in fringe field of spectrometer magnetMeasure x (bending) + stereoextrapolate VELO tracks to TTsplit in two stations ~400 clusters per event

NEW!!

LHCb light

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 15

L1 Algorithm and Decision

Reconstruct r-z tracks and PV with VELOMatch r-z tracks to L0-3D-reconstruct r-z tracks if

0.2 < d/mm < 3, or (why 3 mm ??)belongs to the XX best L0-matched

Determine PT with VELO+TT

Take decision:Cut on log(d/d) % log(PT)

Overrule: if m > 2 GeV then L1-yes

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 16

L1 Trigger performance

Performance improved with addition of TT and L0-matchingFurther improvements with:

IT/OT tracking ?Multiplicity cuts…

VELO onlyVELO only

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 17

FIN

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 18

Trigger performance

L0 (%) L1 HLT Tot µ µµ e γ h all (%) (%) (%)

Bd π+π- 7 6 9 3 55 61 51 ? 31Bs DsK 8 7 5 2 37 44 65 ? 29Bs J/ψ(ee) 7 7 36 4 24 52 43 ? 23Bs J/ψ(µµ) 90 89 5 3 30 93 73 ? 68Bd Κ*γ 6 6 28 47 30 82 33 ? 27

Trigger efficiencies for dimuon channels are ~30% higherHadron trigger is central to LHCb physics goalsEvenly spread selectivity = robustness and flexible

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 19

L0: Muon System High-PTHardware:

Muon System: MWPC, RPCs ?, y-projective, 120k physical channels, 26k, trigger channels, what bits ?

Principle:Muon System: find track, assume origin=IP, use B-kick

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 20

L0: Calorimeter High-PTDetectors:

HCAL / Hadron Calorimeter: Fe & Scint. Tiles, 1468 cells, each 8-bit ET

PS / Preshower: 5952 cells (each 1 bit)SPD / Scintillating Pad Detector: 5952 cells (each 1 bit)ECAL / Electromag. Calorimeter: shashlik, 5952 cells, each 8-bit ET

Principle:ignore B-field, assume origin = IP, use energy and cell position to determine ET (PT)

combine HCAL, ECAL, SPD, PS to produce high-PT

candidates of type electron, gamma, hadron, local pi0, global pi0

polar angle

Energy

2nd Workshop on CKM Unitarity Triangle – April 5-9, 2003, Durham Massimiliano Ferro-Luzzi 21

L1 efficiency depends strongly on SPD cut

egerfer