L1 Track Trigger for ATLAS at HL-LHC Nikos Konstantinidis (University College London) RAL seminar,...

L1 Track Trigger for ATLAS at HL-LHC

Nikos Konstantinidis

(University College London)

RAL seminar, 18/05/2011

Outline

High Luminosity LHC (HL-LHC)Motivation, physics potential, timelines, parameters

ATLAS upgrades for HL-LHCOverview, focus on Level-1 Trigger upgrade

The ATLAS L1Track R&D programmeMotivation, challenges, studies, plans

Conclusions – Outlook

Track Trigger for ATLAS at HL-LHC 2Nikos Konstantinidis

The HL-LHC project

HL-LHC idea

Upgrade the LHC to deliver O(3000fb-1) per experiment in the 2020s (x10 the LHC lumi)O(250fb-1) per year, peak luminosity 5e34

Motivation: Increased discovery potential

multi-TeV region, corners of SUSY parameter space…Detailed studies of discoveries made with 250fb-1

e.g. Higgs couplings, SUSY properties etcProbe the gauge structure of the SM

Vector Boson Fusion at 1TeV, triple gauge couplings…


Timelines for HL-LHC

Current dates for “Long Shutdown 3”: 2021-22Dates driven by two factors:

Integrated lumi “saturation”Radiation damage of accelerator components

Two other long shutdowns in this decade:LS1 (2013-14): work to prepare for 14TeV & 1e34

LHC will have delivered ~O(10fb-1) at 7TeVLS2 (2017/18?): upgrades for peak lumi 2e3

LHC will have delivered ~O(50fb-1) at 14TeV


LHC int. luminosity evolution


If it takes many (>10) years to deliver integrated luminosity that halves statistical errors, then it is not cost-effective.

10fb-1

50fb-1 250fb-1

Track Trigger for ATLAS at HL-LHC 7

LHC: Draft 10-year plan (03/2011)

Nikos Konstantinidis

How to deliver 250fb-1/year

Upgrade the machine to deliver 10e34 (5e34 with luminosity levelingHigher current per bunchHigher field magnets for stronger focusing Superconducting crab cavities for lumi levelingExtreme collimation in the collision points…

All these require a lot

of R&D


Lumi leveling with crab cavities


qc

• RF crab cavity deflects head and tail in opposite direction so that collision is effectively “head on” for luminosity and tune shift

• bunch centroids still cross at an angle (easy separation)• First proposed in 1988, in operation at KEKB since 2007

→ world record luminosity!

Lumi profile with leveling

Much more manageable for the detectors~100 vs. ~200 pp collisions per bunch crossing

Also better for LHCReduced peak heat deposition in critical cold

regions of the machine


0.E+00

2.E+34

4.E+34

6.E+34

8.E+34

1.E+35

0 2 4 6 8 10 12

Lum

inos

ity

(cm

-2s-1

)

time (hours)

Nominal

1035 - no levelling

Levelling at 5 10

35

34

0.E+00

2.E+34

4.E+34

6.E+34

8.E+34

1.E+35

0 5 10 15 20 25

Lum

inos

ity

(cm

-2 s

-1)

time (hours)

1035 - no level Level at 5 1035 34

Average no level

Average level

Example HL-LHC parameters


parameter symbol nom. nom.* HL crab HL sb + lrc HL 50+lrc

protons per bunch Nb [1011] 1.15 1.7 1.78 2.16 3.77

bunch spacing Dt [ns] 25 50 25 25 50

beam current I [A] 0.58 0.43 0.91 1.09 0.95

longitudinal profile Gauss Gauss Gauss Gauss Gauss

rms bunch length sz [cm] 7.55 7.55 7.55 5.0 7.55

beta* at IP1&5 *b [m] 0.55 0.55 0.15 0.15 0.15

full crossing angle qc [mrad] 285 285 (508-622) 508 508

Piwinski parameter =f qcsz/(2*sx*) 0.65 0.65 0.0 1.42 2.14

tune shift DQtot 0.009 0.0136 0.011 0.008 0.010

potential pk luminosity L [1034 cm-2s-1] 1 1.1 10.6 9.0 10.1

events per #ing 19 40 95 95 189

effective lifetime teff [h] 44.9 30 13.9 16.8 14.7

run or level time trun,level [h] 15.2 12.2 4.35 4.29 4.34

e-c heat SEY=1.2 P [W/m] 0.2 0.1 0.4 0.6 0.3

SR+IC heat 4.6-20 K PSR+IC [W/m] 0.32 0.30 0.62 1.30 1.08

IBS e rise time (z, x) tIBS,z/x [h] 59, 102 40, 69 38, 66 8, 33 18, 31

annual luminosity Lint[fb-1] 57 58 300 300 300

ATLAS Upgrades for HL-LHC

ATLAS Upgrades this decade At LS1 (2013-2014)

New Pixel barrel layer very close to the beam (+new beam pipe)The existing innermost Pixel layer

becomes inefficient above 1e34

At LS2 (2017/18?) New Muon inner endcap Wheels

Too high rate in existing chambers above 1e34

New wheels will give sharper L1 muon thresholds in forward region

Trigger upgradesL1 topological processorHardware track finder after L1 (FTK)


ATLAS upgrades for 2021-22 Upgrades are necessary to retain performance at

5e34 or to withstand irradiation up to ~3000fb-1 New, all-silicon tracker (pixels + strips)

New calorimeter readout to provide higher granularity information for the L1 Calo trigger

Major redesign of the L1 Trigger hardware and other upgrades to Trigger-DAQ system

Changes in the very forward calorimeters

Present Calorimeters, Muon Chambers and Magnets don’t change


Current TDAQ architecture

L1 uses only Calo/Muon

40MHz L175kHz Unlikely to change hugely,

given material constraints

L2 accesses only ~10% of event data (in RoIs)

Rate to disk ~200Hz Unlikely to change hugely,

given increased event sizeTrack Trigger for ATLAS at HL-LHC 15Nikos Konstantinidis

L1 Trigger upgrade for 5e34 Physics at HL-LHC will still require triggering on

signatures at the electroweak scale (W/Z/H) Raising the pT thresholds effectively cancels the

benefits of upgrading the LHC luminosity!

L1 Trigger has to achieve 5x higher rejection in much more complex events E.g. electron/muon isolation may be less effective

Reading out the tracker at ~500kHz not an option Costly, wasteful and introduces a lot of materialTrack Trigger for ATLAS at HL-LHC 16Nikos Konstantinidis

Options for ATLAS

Upgrade L1CaloUse finer granularity information from Calorimeters

Upgrade L1Muon Sharpen pT thresholds, reject more fake triggers

Use tracking info at L1Known to be a key element for rejection at L2

A combination of all of the above


L1Calo upgrades for 5e34 Finer granularity info for

L1Calo is the main physics driver for Calo readout upgrade Currently 0.1x0.1 trigger towers

Studies ongoing to understand what info is most important for L1Calo In depth or lateral shower profile Which HLT Calo techniques can

be ported to L1

Strips of Dh=0.003 in 1st LAr sampling are essential for rejecting p0s

ATLAS Trigger upgrade plans and activities 18Nikos Konstantinidis

L1Muon rates at 5e34


Rates for muons from b/c/W can be extracted by convoluting pT resolution with physics x-section

pT resolution drives the physics rates: >50kHz for MU20, >25kHz for MU40 The above estimates do not include cavern background nor charged p/K

decays in flight, hence they give a lower bound

L1Muon upgrades for 5e34 New Small Wheels (to be installed

at LS2) will provide in forward region significant reduction of fake triggers sharper pT thresholds

Barrel L1Muon is harder to improve, but less sensitive to fakes

Ongoing studies Possible L1Mu barrel improvements to characterize fully the L1Muon

performance in MC and project from data to higher luminosities


All L1 Triggers

L1 Triggers with offline muon

Endcaps

Endcaps

The ATLAS L1Track R&D

Why tracking at L1

Tracking is key for rate reduction at L2, so it would enhance the purity of the L1-accepted eventsFor electrons, muons, taus (high-pT tracks)For double/multi-object triggers

(ensure that they come from the same pp collision)

Will provide additional flexibility, redundancy & robustness to the L1 trigger in the harsh and uncertain conditions of HL-LHCOtherwise we’ll have to rely exclusively on pT

thresholds for controlling the L1 ratesTrack Trigger for ATLAS at HL-LHC 22Nikos Konstantinidis

Possible L1Track designs

Self-seeded L1Track: use closely-spaced (few mm) Si layers electrically connected and instrumented with coincidence logic that is satisfied only by high-pT tracks (the CMS idea) Coincidences have to be sent off-detector at 40MHz and

combined in a further step

The L0/L1 idea (or RoI-based L1Track) L1Calo/L1Muon reduce the rate from 40MHz to ~500kHz and

identify Regions of Interest (Level-0) Tracker data from these RoIs are readout and processed by

L1Track; L1Track info is added to the L1Calo/L1Muon info to form global L1 decision Track Trigger for ATLAS at HL-LHC 24Nikos Konstantinidis

Cannot readout the whole tracker at 40MHz

The RoI-based L1Track L0 is a synchronous pipelined trigger with latency ~3.2us

(pretty much like current ATLAS L1 trigger)

Subsystems (e.g. Muons, Calorimeters) can either readout at the full L0 rate (~500kHz), ignoring L1, or have additional buffers (e.g. tracker) in their front-ends where L0-Accapted events are kept until the L1 decision arrives

L0 RoI information is converted into Regional Readout Requests that target specific modules in the tracker

From L0 to L1, the system is asynchronous; makes overall design more flexible

Depending on the size of the additional buffers, the L1 decision can take up to O(100us)Track Trigger for ATLAS at HL-LHC 25Nikos Konstantinidis


Example: strips readout chip design

Advantages of RoI-based L1Track

No impact on the tracker layoutCan be optimized for offline tracking performance

Full tracking information available (in principle)

For larger L1 latency, L1 rate can be significantly below 100kHz Implications for the tracker’s material budget


ATLAS vs. CMS L1 upgrades

Conclusions depend on initial conditionsAt least ~2 years ago, CMS had the constrain for

L1 6.4us/100kHz (no change of ECAL electronics)

ATLAS has more options for improving L1Calo and L1MuonHigh granularity of LAr calorimeter is an advantageMuon Small Wheels

Of course, with more options it is harder to converge


Inner detector

= 3.2 μs

Level-1 Track

ID Tracks

Level-1 MDT

Tracks

Central Topology Processor

L0A

Pre-Trigger Processor

Muon detectors

Level-0 Muon

μ

Calorimeters(Digitisation)

e/γ ETJet

Level-0 Calo

R3

Trigger & Timing distribution

L0A, L1A

L1A

25 ns perpipeline step

= < 256 - 3.2 μs

A possible architecture: L0/L1

40M

Hz

~50

0kH

z?<

100k

Hz?

Nikos Konstantinidis 29Track Trigger for ATLAS at HL-LHC

RoI-based L1Track parameters

Key parameters for the design of L1Track areL0 rateL1 rate

could be <100kHz if enough rejection is achievedNumber/size of RoIs

hence the fraction of the tracker to readout at L0 rateNumber of tracker layers to readout

not necessarily all

Ongoing simulation studies to determine these parameters


Regional Readout studies

Track Trigger studies 31Nikos Konstantinidis

Mod

ule

num

ber

in Z

Frequency (in %) inside an RoI

PIXEL

STRIPS

RoI: Df=0.2, Dh=0.2 at Calo Dz=40cm at beamline

RoI-based L1Track – issues All ATLAS sub-detector should either be able to

readout at the L0 rate (~500kHz) or change their readout electronics Not an issue for tracker (new), Calorimeters (new

electronics) and other small sub-detectors BUT: potential problem with Muon system: max readout rate

100kHz (max latency 3.2us) and replacing all electronics is a very long operation – unlikely to fit in a 2-year shutdown

Alternatives are under investigation

Unlikely that any L1Track scheme can fit within 3.2us/100kHz


Conclusions – Outlook The HL-LHC will dominate the high energy discovery

frontier in the 2020s

Major upgrades of the detectors will be required to cope with the high-lumi environment (up to 200 pp collisions / bunch crossing)

ATLAS is evaluating a number of alternatives for the upgrades of L1Calo and L1Muon, as well as the use of tracking information at L1

An RoI-based L1Track appears to be an attractive solution for ATLAS, but more studies are needed to converge to the optimal L1 architecture Track Trigger for ATLAS at HL-LHC 33Nikos Konstantinidis

L1 Track Trigger for ATLAS at HL-LHC Nikos Konstantinidis (University College London) RAL seminar,...

Documents

Transcript of L1 Track Trigger for ATLAS at HL-LHC Nikos Konstantinidis (University College London) RAL seminar,...