The ATLAS Trigger

Teresa Fonseca Martín (RHUL)

RHUL Special Seminar

April 16 2008

2

Outline• Motivation• Few words on LHC• ATLAS trigger requirements• Basis of ATLAS trigger• Atlas detector picture• L1 Trigger

– Ex: L1 electromagnetic trigger

• High Level Trigger (HLT)– Ex: e

Some relevant open questions

1. Mass:Mass: What is the origin of mass?

- How is the electroweak symmetry broken ?

- Does the Higgs boson exist ?

2. Unification:Unification: What is the underlying fundamental theory ?

Motivation: Gravity not yet included; Standard Model as a low energy approximation

- Is our world supersymmetric ?

- Are there extra space time dimensions ?

- Other extensions ?

3. Flavour:Flavour: or the generation problem

- Why are there three families of matter?

- Neutrino masses and mixing?

- What is the origin of CP violation?

The role of Hadron Colliders 1. M1. Massass

- Search for the Higgs boson

2. Unification2. Unification

- Test of the Standard Model

- Search for Supersymmetry

- Possible link between SUSY and Dark Matter?

- Search for other Physics Beyond the SM

3. 3. FlavourFlavour

- B hadron masses and lifetimes

- Mixing of neutral B mesons

- CP violation Energy Explore the TeV energy domain Experiments must also be prepared for “the unexpected”

Precision Further tests of the Standard Model

5

The Large Hadron Collider

LHCb

CMS

ALICEATLA

S

proton-proton collisions at an energy of 14 TeV

26.7 km circumference

start: Fall 2008

6

Some LHC parameters• Bunch spacing is 25 ns

• Bunch-crossing rate is 40 MHz

• Total proton-proton cross-section ~ 100 mb– Interaction rate at L = 1034 cm-2s-1 is R ~ 109 Hz

• On average ~ 25 interactions per bunch crossing – Background activity that complicates analysis of what happened

in the interaction of interest

• Centre-of-mass energy s = 14 TeV for proton-proton collisions

• c.f. 2 TeV at Tevatron collider• Foreseen start-up: s ~ 10 TeV

• Luminosity– L = 1034 cm-2s-1 (for proton-proton collisions in ATLAS and CMS)

• c.f. L = 1032 cm-2s-1 at Tevatron• Foreseen start-up: L~1031 cm-2s-1

7

µµnn

pp

ee

Proton-Proton Collisions at the LHC

p pH

µ+

µ-

µ+

µ-

Z

Zp p

e- e

q

q

q

q

1

-

g~

~

20~

q~

10

~

7 TeV Proton Energy1034cm-2s-1 Luminosity2835 Bunches per Beam1011 Protons per Bunch

7.5 m 25ns

Bunch-Crossings 4x107 Hz

Proton-Proton Collisions 109 Hz

Quark/Gluon Collisions

Heavy particle production 10+3…-7 Hz(W, Z, t, Higgs, SUSY, BH,…)

8

ATLAS triggerWe can not afford to record all the LHC collisions• The trigger is the system that selects particle interactions

that are potentially of interest for physics analysis

• The ATLAS trigger has to provide the data for:– Standard Model precision measurements and searches for new

physics– Physics processes needed for commissioning, calibration,

background studies, monitoring, etc

• The ATLAS trigger has to have a high rejection power to keep the rates within the allocated bandwidth

• The trigger has to be fast and robust– E.g. robust against pile-up, misscalibration,misalignment, etc

9

Data that trigger should provide (I)

• Typical examples of new physics that trigger selection should cover:– Origin of electro-weak symmetry breaking (mW and mZ)

• Higgs boson (Standard Model and beyond)• Alternative schemes

– SUSY• With and without R-parity conservation

– Compositeness– Leptoquarks– W’ and Z’– Extra dimensions

• KK excitations, black holes

– The unpredicted!– LHC has an order of magnitude more centre-of-mass energy and

two orders of magnitude more luminosity compared to today’s most powerful machine

10

Data that trigger should provide (II)

• Standard Model production processes– W, Z, direct-photon production– Jet production (including multi-jet production)

• Interesting in their own right• Must be understood as backgrounds to new physics

• B-physics studies (especially in the early phases) • Standard Model precision measurements

– W mass and top mass• Important for consistency checks with Higgs studies

11

LHC environment• The trigger needs to reduce the event rate to a

manageable level for data recording and offline analysis– L = 1034 cm-2s-1, and ~ 100 mb 109 Hz interaction rate

• Even rate of events containing leptonic W and Z decays is O(100 Hz)

– The size of the events is very large, O(1) MByte• Huge number of detector channels, high particle multiplicity per

event

– Recording and subsequently processing offline, O(100) Hz event rate with O(1) MByte event size implies major computing resources!

– Hence, only a tiny fraction of proton–proton collisions can be selected

• Maximum fraction of interactions triggering at full luminosity O(10-7)

12

Multiple interactions per BC (“pileup”)

Higgs → ZZ → 2e+2Higgs → ZZ → 2e+2 23 min bias events/bunch crossing~1725 particles produced

23 min bias events/bunch crossing~1725 particles produced

13

Trigger requirements summary

• Discovery physics– Vast range of predicted

processes with diverse signatures

• Very low signal rates expected in some cases

– Sensitive to unpredicted new physics

• Huge rate of Standard Model physics backgrounds

• Balance: maximising physics coverage and reaching affordable recording rates

14

ATLAS trigger• L1

– Hardware based– < 2.5s. Output rate: ~100KHz– Coarse granularity– Calorimeter and muon syst.

• L2– Software based– ~40ms. Output rate: ~1kHz– RoI seeded by L1– Full detector granularity – Fast tracking and calorimetry

• EF– “Offline-like” algorithms– ~4s . Output rate: ~200Hz– RoI seeded by L2 (full detector possible)– Full aligment/calibration info

15

Trigger signatures

Trigger objects•Electron/Photon•Tau•Jets•Muon•B-Physics•Missing ET

•b-tagging•Minimum bias

e

jet

IDET ECAL HCAL MuDET

16

The ATLAS detector

17

Trigger menus

• Different threshold values in selection cuts can be applied.– Ex:

• e15 electron ET > 15GeV• e15i: isolated e ET > 15GeV• e60: electron ET > 60GeV

• Different objects combined:– Ex: e15i+Missing ET

– Ongoing work to define combined signatures menu

Trigger objects•Electron/Photon•Tau•Jets•Muon•B-Physics•Missing ET

•b-tagging•Minimum bias

Remember you want to be sensitive to the unpredicted: be as inclusive as you can

18

ATLAS LVL1 Trigger

Calorimeter trigger Muon trigger

Central Trigger Processor (CTP)

Timing, Trigger, Control (TTC)

Cluster Processor (e/, /h)

Pre-Processor (analogue ET)

Jet / Energy-sum Processor

Muon Barrel Trigger

Muon End-cap Trigger

Muon-CTP Interface (MUCTPI)

Multiplicities of for 6 pT thresholdsMultiplicities of e//h, jet

for 8 pT thresholds each; flags for ET, ET

jet, ETmiss over

thresholds; multiplicity of fwd jets

LVL1 Accept, clock, trigger-type to Front End systems, RODs, etc – RoI pointers

~7000 calorimeter trigger towers O(1M) RPC/TGC channels

ET values (0.20.2)EM & HAD

ET values (0.10.1)EM & HAD

pT, information onup to 2 candidates/sector(208 sectors in total)

19

L1 e trigger

• ATLAS e/ trigger is based on 44 “overlapping, sliding windows” of trigger towers– Each trigger tower 0.10.1

in

• pseudo-rapidity, azimuth

– ~3500 such towers in each of the EM and hadronic calorimeters

• 8 sets of thresholds(+ 8 shared with )

20

ATLAS LVL1 μ trigger

• ATLAS LVL1 μ trigger is based on coincidences among hits within “window” in layers of RPCs (TGCs).

Window size determines pT threshold.– Low pT trigger 3 thresholds

– High pT trigger 3 thresholds

21

High Level Trigger (HLT)

• Software based:– L2: Specific fast trigger algorithms– EF: Uses offline algorithms as much as

possible

• Key ATLAS trigger design concepts:– Region of Interest (RoI) mechanism– Early rejection

22

Region of Interest (RoI)

• L1 indicates the geographical location of candidate objects (,)

• L2 only access data from a detector subregion around (,): “Region of Interest” (RoI)

• Reduces L2 network bandwith• Reduces L2/EF processing

time

23

HLT general scheme• “Trigger chain”: each trigger signature (ex: e25i)

corresponds to a chain of algorithms that are executed sequentially– 2 types of algorithms:

• Feature extraction algorithms (FEX): create EDM objects (clusters, tracks, etc)

• Hypothesis algorithms: apply selection cuts

• “Trigger steering”: master algorithm that controls the HLT run– According to the loaded configuration (menu) establishes chains to be

executed– Calls sequentially the algorithms that have to be run– After each “Hypo” step verifies if trigger element is active

– Early rejection: As soon as a trigger requirement is not satisfied the execution of the corresponding chain is stopped

24

Electronslice

T2CaloL2CaloHypoL2Tracking

L2CaloIDHypo

TrigCaloRecEF ID

TrigEgammaRec

EFTrackHypo

EFEgammaHypo

L2 Calo+Trk Match

T2Calo

L2CaloHypo

TrigCaloRec

TrigEgammaRec

EFEgammaHypo

L2Photon FEX

L1L1 PhotonsliceTrigEMCluster

TrigPhoton

egamma

TrigElectron

CaloClusterCaloCell

TrigInDetTrack

TrackParticle

•FEX algorithms: create EDM objects•Hypothesis alg.: apply selection cuts

L2: specific trigger algorithms EF: use of offline tools as possible

Example of a slice: e slice

25

e L2 CaloT2CaloEgamma:

• Performs calorimeter cluster reconstrunction.

• Full detector granularity• Builds shower shape

variables to discriminate electron/photon of jets

L2 Calo Hypo:• Applies selection cuts on

shower shape to discriminate electron/photon of jets

Current L2 calo selection:

26

• Look for additional maximum in sampling 1 (E1-E2)/(E1+E2). (~1 for electrons)

Ex. of L2 calo selection variable

0

E1

E2

27

e L2 tracking + calo combinationCurrently two alternative tracking algorithms implemented: IDSCAN:

• zFinder: Reconstruction of the z-position of the primary pp collision• hitFilter & groupCleaner: The main pattern recognition step• trackFitter: final track fit and removal of outliers

SiTrack:• Space point sorting• Track seeds formation• Primary vertex reconstruction• Track extension

TRT standalone is being improved

Cluster-Track matching to build L2 electron object

• match• match• ET/pT (currently not active)

28

e Event FilterTrigCaloRec:

• Performs calorimeter cluster reconstruction• Wraps-up offline tools • Involved also in the tau and jet slices

EFID:• Based on offline tools in a seeded mode• Involved in the tau, b-physics, b-tagging and

muon slices also TrigEgammaRec

• Reconstructs the EDM egamma object• Wraps-up offline tools • Combines Inner Detector and Calorimeter

information• Includes possibility of bremstrahlung correction

e EF selection: currently uses isem (offline identification)For identification not ID information used

ET/pT without brem recoverywith brem recovery

29

Examples of electron slice performance studies

• Study trigger efficiency dependencies on individual cuts and ET, and .

• Compare electrons from single electron and from Zee, + pile-up effects

Ex: e25i signature

• Scan selection cuts thresholds• For a given rate maximum trigger efficiency

Example of L2 selection optimization

30

Developing methods to determine trigger efficiency from data (Zee)

• Control sample: reconstruct Z + 1e trigger

• Determine trigger efficiency checking if second electron has been triggered

Determine differential trigger efficiency (vs , and ET)

2 artificial inefficient regions in for L1

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

Cannot rely on Monte Carlo models, especially in early days Very important to develop tools to measure trigger efficiency from data with as little reliance as possible on MC

31

Selection slices• Electron/Photon• Jets• Tau• Muon• B-Physics• Missing ET

• b-tagging• Minimum bias• Cosmics

Require detailed detector understanding, may not be available on day-1

Comparatively very low pT threshold, take advantage of low-luminosity era

Already started

Crucial for large part of physics programme and must be operational from the start

Too long to cover it in detailJust give the flavor with a few examples

32

Tau slice

95% efficiency

lowest rate

Ex. L2 optimization

EF turn on curves

combined run Maymean 6.2 ms

tau 15itau 20itau 25itau 35i

33

Jet SliceRetrieve cells in the RoI Fast cone algorithm (R=0.4)Calibrate the jet

ET cut

“Offline” jet cone (or fastKt) reconstruction - Uses offline tools and calibration

Calorimeter towers (or topo cluster)

Data unpacking

ET cut

34

ROIROI

Level-2 jet trigger• Level-1 RoI is passed to Level-2Level-1 RoI is passed to Level-2

• LVL2: LVL2: iterative (3 iter.) cone iterative (3 iter.) cone

algorithm calculates energy-algorithm calculates energy-weighted position (weighted position (,,..

Apply simple, robust, fast Apply simple, robust, fast calibration procedurecalibration procedure..

Half Width

L2 jetL2 jet

Niterationgrid element

ConeRadius

Main difficulty: jet energy scale calibration

•JES within 2% correct•Resolution fit to

35

Muon slice

L2:• muFast: muon spectrometer stand alone reconstruction (η, φ and pT)

• Track reconstruction efficiency: ~99.5% barrel, ~100% endcap• muComb: refines muon tracks combining them with the Inner Detector

track.• muIso: Calorimeter isolation algorithm to reject muons from beauty and

charm semileptonic decays.EF:• Wraps offline reconstruction

L2 EF

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.

36

Examples of muon slice performance studiesExtensive studies of efficiency and resolution for different thresholds, regions, micalibration and misalingment

6GeV

40GeV

8GeV

20GeV

Barrel EF MuId combined

Barrel

muComb

EC L2 resolutionComb recoversthe resolutionIn all regions

37

B-Physics slice

Grid samplephi massall eta values

Mean=1019.5Sigma = 4.92

First studies with pileup No pileup Pileup

L1_MU06 81 +/- 2 % 91 +/- 1.5 %

+ L2_DsPhiPi_RoI 68 +/- 2 % 79 +/- 2 %

+ L2_DsPhiPi_FullScan 74 +/- 2 % 79 +/- 2 %

J. Kirk

Triggers currently being studies & developed for:

– B->J/ XJeeXX

– B-> B->K*B->– B->K*B->– Bs->Ds, Bs->Dsa

– B->J/X, B->K*/ etc

• B-Physics strategies to keep rate within budget:• Low luminosity: lowest

possible muon pT threshold and full InnerDetector reconstruction

• High luminosity: higher pT threshold (mu6) and RoI based reconstruction

Example: Bs→Ds(Φπ)π L2 trigger

38

Missing ET

– RoI concept does not apply to global quantity– Data preparation is a major concern when accessing entire calorimeter

• L2– L1 Missing ET + all L2 Muons

• EF– default Algorithm = loop over all cells at EM-scale– alternative algorithm = loop over Ex/Ey sums in FEB header– + muons– simple hadronic calibration

Ex. resolution studies:

39

b-tagging

L1: use jet thresholds

HLT:• 3j/2b or 4j/3b

– b-tagging 70%eff.• (Offline b-tag 60%)

• L2 tracking• EF tracking• Hypo: likelihood based on impact parameter

• Under study:– Use cluster to get jet direction– Use more offline “tools”, ex.:

secondary vertex

Significance of longitudinal impact parameter

Rejection for u-jets vs b-tag trigger efficiency

40

Summary

• ATLAS trigger challenges have been summarized:– Need high efficiency for selecting processes for physics

analysis• Selection should not have biases that affect physics results

– Need large reduction of rate (~O(1/107)) from unwanted high-rate processes

– System must be affordable and robust• e.g algorithms executed at high rate must be fast

• An overview of Atlas trigger system has been given– L1 & HLT have been outlined

• As an example e trigger has been explained wirh more detail

Spares

42

LHC uses existing CERN complex

• LHC is being built in the existing tunnel previously used for LEP

– Circumference = 27 km• Radius = 4.3 km

• Use existing accelerators as injection system

• Four major “experiments”– ATLAS and CMS are “general-

purpose” detectors optimised for exploring new physics in pp collisions

– LHCb is a specialized detector optimised for B-physics studies

– ALICE is a specialized detector optimised for heavy-ion physics

4343Chris Bee - CPPM – April 2007Chris Bee - CPPM – April 2007

LVL1

ATLAS Trigger/DAQATLAS Trigger/DAQATLAS Trigger/DAQATLAS Trigger/DAQ

40 MHz

75 kHz

~3 kHz

~ 200 Hz

120 GB/s

~ 300 MB/s

~2+4 GB/s

Trigger DAQ

2.5

s

Calo MuTrChOther detectors

FE Pipelines

RoI

Lvl1 acc = 75 kHz

40 MHz

Read-Out Sub-systems

Event Building N/work

DATAFLOW

EB

ROSH

L

T

LVL2~ 10 ms

SFI

SFO

EBN

EFN

Read-Out Drivers

Dataflow Manager

Sub-Farm Input

Event Filter N/work

ROIB

L2P

L2SV

L2N

Event Filter

DFM

EFPEFP

EFPEFP

RoI Builder

L2 SupervisorL2 N/work

L2 Proc Unit

RoI data = 1-2%

RoI requests

Lvl2 acc = ~3 kHz

~ sec

120 GB/s

~4

GB

/s

EFacc = ~0.2 kHz

Read-Out Buffers

Read-Out Links

Event Builder

DE

T ROD ROD ROD

ROB ROB ROB

44

High empty event rejection while high trigger efficiency for minbias can be obtained by cutting on SCT spacepoints

Track multiplicity cut at 5 tracks gives 93% minbias trigger efficiency while beamgas events are strongly supressed

Minimum Bias Trigger

R. Kwee

• Minimum bias trigger strategies:– Minimum bias trigger

scintillators (MBTS)– Cut on number of SCT

space points– Cut on number of tracks

R. Kwee

E. Feng

MBTS

45

Accidental Rate

• Noise spectrum fits well to a Gaussian with σ=2.52pC (left plot)• Beyond 9pC (almost 4σ), non-Gaussian behavior is possible• Accidental rate must be suppressed to ~Hz, as Event Filter can

only write 100Hz

46

MBTS Counters

• “Inner” counters: 2.83 ≤ η ≤ 3.85• “Outer” counters: 2.12 ≤ η ≤ 2.83• 8 inner, 8 outer on each of A and

C sides 32 total counters

47

Data Flow for the HLTLVL2Supervisor

Database

ROS

L2PU

Event Building

SFI/SFO

EFD

LVL1Result

LVL2Decision

LVL2Selection

EFSelection

DataRequest

EventFragments

EventFragments

DataRequest

LVL1 Result

LVL2 Decision

FullEvent

FullEvent

Full Event+ EF Result

SelectedFull Event

LVL2 Result

48

Particles cartoon

e

jet

IDET ECAL HCAL MuDET

49

Table rates and tr item

50

Global Requirements of TDAQ Systems at LHC

Rejection power for Higgs search 1013

No overlapping events / 25 ns 23 No particles in ATLAS / 25 ns 1400 Data throughput

At detectors (40 MHz) (equivalent to) PB/s

--> LVL1 Accepts 100 GB/s --> Mass storage 100 MB/s

51

Some LVL1-trigger design goals

• Need large reduction in physics rate already at the first level (otherwise readout system becomes unaffordable)– O(109) interaction rate less than 100 kHz in ATLAS and CMS

• Require complex algorithms to reject background while keeping signal

• An important constraint is to achieve a short latency– Information from all detector channels (O(108) channels!) has to be held

in local memory on detector pending the LVL1 decision• The pipeline memories are typically implemented in ASICs (Application

Specific Integrated Circuits), and memory size contributes to the cost

– Typical values are a few s (e.g. less than 2.5 s ATLAS, 3.2 s CMS)

• Require flexibility to react to changing conditions (e.g. wide luminosity range) and — hopefully — new physics– Algorithms must be programmable (adjustable parameters at least)

52

Inclusive Selection Signatures

Object Examples of physics coverage Nomenclature

ElectronsHiggs (SM, MSSM), new gauge bosons,

extra dimensions, SUSY, W, top e25i, 2e15i

Photons Higgs (SM, MSSM), extra dimensions, SUSY 60i, 220i

MuonsHiggs (SM, MSSM), new gauge bosons,

extra dimensions, SUSY, W, top 20i, 210

Jets SUSY, compositeness, resonances j360, 3j150, 4j100

Jet+missing ET SUSY, leptoquarks j60 + xE60

Tau+missing ET Extended Higgs models (e.g. MSSM), SUSY 30 + xE40_ __

ATLAS is a multipurpose experiment aiming at discovery and precision measurements of “expected” and “unexpected” Physics signals.

The selection of Physics signals requires the identification of objects that can be isolated from the high particle density environment.

The list must be flexible, extendable, non-biasing and general enough to account for the “unexpected”.

53

LEP requirements (1)

• Trigger had to:– Identify all events coming from e+e- annihilations with visible final

states• Including at LEP-I: Z hadrons, Z e+e-, Z +-, Z +-

• Including at LEP-II: WW, ZZ, single-boson

• Including cases where there is little visible energy

– e.g. in Standard Model: e+e- Z – e.g. in new particle searches such as e+e- (with small – mass difference), giving only low energy visible particles ( LSP)

– Retain some fraction of two-photon collision events• Used for QCD studies

– Identify Bhabha scatters • Needed for precise luminosity determination

54

LEP requirements (2)

• Could retain events with any significant activity in the detector– Even when running at Z peak, rate of Z decays was only O(1 Hz)

• Physics rate was not an issue

• Challenge was in maximising efficiency/acceptance of trigger– And also, making sure that the efficiency and acceptance could be

determined with very high precision• Absolute cross-section determination depends on knowing:

– Integrated luminosity (efficiency to trigger on Bhabha events)– Experimental efficiency and acceptance for process in question

(efficiency to trigger on physics process)» Events selected by many redundant triggers

(high efficiency; cross-checks)• A major achievement at LEP was to reach per-mil precision

• The trigger rates and also the DAQ data rates were modest

55

Path of the CERN LEP/LHC tunnel

CERN

Airport

Circumferenceof ring ~ 27 km

56

LVL1 signatures at hadron colliders• LVL1 triggers therefore search for

– High-pT muons• Identified beyond calorimeters; need pT cut to control rate from ,

K+ , as well as semi-leptonic beauty and charm decays

– High-pT photons• Identified as narrow EM calorimeter clusters; need cut on ET; cuts on isolation

and hadronic-energy veto reduce strongly rates from high-pT jets

– High-pT electrons• Same as photon (some experiments require matching track already at LVL1)

– High-pT taus (decaying to hadrons)• Identified as narrow cluster in EM+hadronic calorimeters

– High-pT jets• Identified as cluster in EM+hadronic calorimeter — need to cut at very high pT

to control rate (jets are dominant high-pT process)

– Large missing ET or total scalar ET

57

LVL1 selection criteria• Features that distinguish new physics from the bulk of the cross-

section for Standard Model processes at hadron colliders are– In general, the presence of high-pT particles (or jets)

• e.g. these may be the products of the decays of new heavy particles– In contrast, most of the particles produced in minimum-bias interactions

are soft (pT ~ 1 GeV or less)

– More specifically, the presence of high-pT charged leptons (e, ), photons and/or neutrinos

• e.g. the products (directly or indirectly) of new heavy particles– These give a clean signature c.f. low-pT hadrons in minimum-bias case,

especially if they are “isolated” (i.e. not inside jets)

– The presence of known heavy particles• e.g. W and Z bosons may be produced in Higgs particle decays

– Leptonic W and Z decays give a very clean signature» Also interesting for physics analysis and detector studies