Trigger Strategies for Long-Lived Particles LHC Searches for LL BSM - - PowerPoint PPT Presentation

SLIDE 1

Trigger Strategies for Long-Lived Particles

LHC Searches for LL BSM Particles: Theory Meets Experiment UMass - 12-14 Nov 2015

Andrea Coccaro

University of Geneva

Andrea Coccaro 1 14 November, 2015 - Trigger Strategies for LLP

SLIDE 2

Introduction

Need to be smart in building the triggers and in picking up the right triggers

• 1. dedicated vs standard
• 2. inclusive vs exclusive
• 3. threshold vs multiplicity
• 4. signal vs control regions
• 5. ...

The trigger-building primer

◮ function of mass, decay products, boost, lifetime ◮ rough information at L1, no tracks ◮ limited readout and limited reconstruction at HLT ◮ early rejection principle ◮ driven by rate reduction, then by signal efficiency

Trigger strategies are evolving fast, but targets are moving as well!

Andrea Coccaro 2 14 November, 2015 - Trigger Strategies for LLP

SLIDE 3

Outline

• 1. LLP triggers in calorimeter and muon spectrometer
• 2. LLP triggers in the tracker
• 3. other dedicated triggers
• 4. simplified menu evolution
• 5. exotic approaches (?)

Andrea Coccaro 3 14 November, 2015 - Trigger Strategies for LLP

SLIDE 4

Calorimeter Ratio Trigger

Ingredients:

◮ TAU40 at L1 ◮ track and jet reconstruction at the HLT ◮ no tracks with pT > 1 GeV in ∆R < 0.2 around the jet axis ◮ log(EHAD/EEM) > 1.2 ◮ beam halo removal using calorimeter cell timing

Energy [a. u.] 1 10

2

10

3

10

4

10

5

10

6

10

7

10

jet

φ

• cell

φ

• 0.4
• 0.3
• 0.2
• 0.1

0.1 0.2 0.3 0.4

jet

η

• cell

η

• 0.4
• 0.3
• 0.2
• 0.1

0.1 0.2 0.3 0.4 ATLAS Simulation =20 GeV

V

π

=140 GeV, m

H

m

r [m] 0.5 1 1.5 2 2.5 3 3.5 4 〉 )

EM

/E

HAD

(E

10

log 〈

• 1
• 0.5

0.5 1 1.5 2 ATLAS Simulation =20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m

Key feature: Isolated jet with very low EM fraction

Andrea Coccaro 4 14 November, 2015 - Trigger Strategies for LLP

SLIDE 5

Muon RoI Cluster Trigger

Ingredients:

◮ 2MU10 at L1 ◮ muon cluster asking for 3 (4) muon RoIs in ∆R < 0.4 in the barrel (end-cap) MS ◮ no tracks with pT > 5 GeV in ∆R < 0.4 around the muon cluster direction ◮ no jets with ET > 30 GeV in ∆R < 0.7 around the muon cluster direction

r [m] 1 2 3 4 5 6 7 8 9 10 Average number of RoIs 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m

|z| [m] 2 4 6 8 10 12 14 Average number of RoIs 1 2 3 4 5 6 7 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m

Key feature: Isolated cluster of muon region of interests

Andrea Coccaro 5 14 November, 2015 - Trigger Strategies for LLP

SLIDE 6

Timing window acceptance

Jet energy within 1% for time shifts up to 12 ns - reference

◮ ≥ 93% of the πv decays arrive in the HCal with a ∆t < 10 ns

Muon trigger accurately matches events for ∆t < 6 ns - reference

◮ ≥ 75% of the πv decays arrive in the MS with a ∆t < 6 ns

mean proper lifetime [m]

v

π 5 10 15 20 25 30 35 t < 10 ns ∆ decays with

v

π Fraction of 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m

mean proper lifetime [m]

v

π 5 10 15 20 25 30 35 t < 6 ns ∆ decays with

v

π Fraction of 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m

Andrea Coccaro 6 14 November, 2015 - Trigger Strategies for LLP

SLIDE 7

Trigger Efficiency

(with respect to truth)

r [m] 1 2 3 4 5 6 7 8 9 10 Trigger Efficiency 0.05 0.1 0.15 0.2 0.25 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m

r [m] 1 2 3 4 5 6 7 8 9 10 Trigger Efficiency 0.1 0.2 0.3 0.4 0.5 0.6 0.7 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m Andrea Coccaro 7 14 November, 2015 - Trigger Strategies for LLP

SLIDE 8

LLP triggers in action

Triggers for displaced decays of long-lived neutral particles [JINST 8 (2013) P07015] Standalone vertex finding in the ATLAS muon spectrometer [JINST 9 (2014) P02001] Search for a light Higgs boson decaying to LL particles [PRL 108 (2012) 251801] Search for long-lived neutral particles decaying into lepton jets [JHEP 11 (2014) 088] Search for pair-produced LL neutral particles decaying in the HCal [PLB 743 (2015) 15-34] Search for LL particles that decay to displaced hadronic jets [Phys. Rev. D 92, 012010 (2015)]

[GeV]

d

γ

m

• 3

10

• 2

10

• 1

10 1 ∈ Kinetic mixing parameter

• 11

10

• 10

10

• 9

10

• 8

10

• 7

10

• 6

10

• 5

10

• 4

10

• 3

10

• 2

10 ATLAS

= 8 TeV s

• 1

20.3 fb

SN LSND BaBar A1 APEX KLOE HADES CHARM E137 U70 Orsay E141 E774

e

a

σ ,5 µ

a favoured

σ 2 ± , µ

a

90% CL ATLAS

BR 40% BR 20% BR 10% BR 5%

) [m] τ c proper lifetime (

v

π

1 −

10 1 10

2

10

SM

σ BR / × σ 95% CL Upper Limit on

2 −

10

1 −

10 1 ATLAS

• 1

= 8 TeV, 19.5 fb s

= 125 GeV

H

m

=10 GeV

V

π

m =25 GeV

V

π

m =40 GeV

V

π

m

BR 15% BR 5% BR 1%

) [m] τ c Singlino proper lifetime (

1 −

10 1 10

2

10 BR [pb] × σ 95% CL Upper Limit on

2 −

10

1 −

10 1 10

2

10 ATLAS

• 1

= 8 TeV, 19.5 fb s =110 GeV

g ~

m =250 GeV

g ~

m =500 GeV

g ~

m =800 GeV

g ~

m =1200 GeV

g ~

m

Andrea Coccaro 8 14 November, 2015 - Trigger Strategies for LLP

SLIDE 9

Outline

• 1. LLP triggers in calorimeter and muon spectrometer
• 2. LLP triggers in the tracker
• 3. other dedicated triggers
• 4. simplified menu evolution
• 5. exotic approaches (?)

Andrea Coccaro 9 14 November, 2015 - Trigger Strategies for LLP

SLIDE 10

Trackless Jet Trigger

Ingredients:

◮ MU10 and J20 at L1 ◮ track, jet and muon reconstruction at the HLT ◮ confirmation of muon and jet at L1 within ∆R < 0.4 ◮ no tracks with pT > 0.8 GeV in ∆R < 0.2 around the jet axis

Number of L2 tracks 2 4 6 8 10 decays

v

π Fraction of 0.1 0.2 0.3 0.4 0.5 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m

r [m] 1 2 3 4 5 6 7 8 9 10 Trigger Efficiency 0.01 0.02 0.03 0.04 0.05 0.06 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m =10 GeV

V

π

=126 GeV, m

H

m =25 GeV

V

π

=100 GeV, m

H

m

Key feature: Jet (and muon) with track isolation

Andrea Coccaro 10 14 November, 2015 - Trigger Strategies for LLP

SLIDE 11

Expected Fraction of Events

mean proper lifetime [m]

v

π

• 1

10 1 10 decays

v

π Fraction of triggered 0.02 0.04 0.06 0.08 0.1 0.12 ATLAS Simulation

=20 GeV

V

π

=140 GeV, m

H

m Trackless Jet trigger Calorimeter Ratio trigger Muon RoI Cluster trigger

Nice complementarity among the three triggers

Andrea Coccaro 11 14 November, 2015 - Trigger Strategies for LLP

SLIDE 12

Trigger Rates and Pile-Up Dependence

]

• 1

s

• 2

cm

33

• Inst. Luminosity [10

2 2.5 3 3.5 4 4.5 5 5.5 Rate [Hz] 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 ATLAS

26.6 〉 µ 〈 = 8 TeV, peak s

Trackless Jet trigger Calorimeter Ratio trigger Muon RoI Cluster trigger

〉 µ 〈 5 10 15 20 25 30 35 40 Ratio of trigger rates 0.2 0.4 0.6 0.8 1 1.2 ATLAS Simulation

>5 GeV

T

>3 GeV / p

T

p >5 GeV

T

>1 GeV / p

T

p

The track-isolation requirement on the Muon RoI Cluster trigger (pT > 5 GeV) is not pile-up dependent while requiring track isolation for tracks with pT > 1 GeV makes the trigger response pile-up dependent.

Andrea Coccaro 12 14 November, 2015 - Trigger Strategies for LLP

SLIDE 13

25 reconstructed events in ∼ 5 cm in a Z → µµ candidate event (2012 data)

Andrea Coccaro 13 14 November, 2015 - Trigger Strategies for LLP

SLIDE 14

CMS Displaced Jet Trigger

Targeting decays within the tracking volume

◮ HT > 300 GeV using all jets with ET > 40 GeV and |η| < 3 ◮ at least 2 jets with ET > 60 GeV with no more than

◮ 2 tracks with 3D impact parameter < 300µm ◮ 15% of the jet total energy associated to tracks with 3D impact parameter < 500µm

Non sensitive to the Higgs at 125 GeV but greatly sensitive to short lifetimes

CMS

(8 TeV)

• 1

18.5 fb

[cm] τ X c 1 10 ) [pb] q q → (X

2

B × XX) → (H σ

• 1

10 1 10 95% CL limits: = 50 GeV

X

m ) σ 1 ±

• Exp. limits (

= 200 GeV

H

m CMS (8 TeV)

• 1

18.5 fb

[cm] τ X c 1 10

2

10 ) [pb] q q → (X

2

B × XX) → (H σ

• 3

10

• 2

10

• 1

10 1 95% CL limits: = 50 GeV

X

m = 150 GeV

X

m ) σ 1 ±

• Exp. limits (

= 400 GeV

H

m

) [m] τ c proper lifetime (

v

π

1 −

10 1 10

2

10 BR [pb] × σ 95% CL Upper Limit on

2 −

10

1 −

10 1 10

2

10 ATLAS

• 1

= 8 TeV, 19.5 fb s =50 GeV

V

π

=300 GeV, m

Φ

m =50 GeV

V

π

=600 GeV, m

Φ

m =150 GeV

V

π

=600 GeV, m

Φ

m =50 GeV

V

π

=900 GeV, m

Φ

m =150 GeV

V

π

=900 GeV, m

Φ

m

Andrea Coccaro 14 14 November, 2015 - Trigger Strategies for LLP

SLIDE 15

LLP trigger in the tracker

b-Tagging is NOT efficient for LLP

◮ tracks with large impact parameters are expressly vetoed ◮ algorithms highly tuned against b-jets versus light jets ◮ but the bulk of the time is spent in getting the tracks

◮ tracking is typically run in events with low-pT jets ◮ some tracks from the LLP will have a not too large impact parameter ◮ dedicated selection?

Explicit DV reconstruction

◮ requires efficient tracking at large impact parameters ◮ not yet there

No actual DV reconstruction

◮ trackless jet allowing soft-tracks from pile-up ◮ tracks with a lower total momentum compared to the jet one ◮ larger hit-occupancy / track ratio as a function of the modules

Can we do better? Most probably YES, studies are needed

Andrea Coccaro 15 14 November, 2015 - Trigger Strategies for LLP

SLIDE 16

Tracking Timing Improvements at the HLT

Tracking in two steps in Run-II

Total Time per Event [ms] 200 400 600 800 1000 1200 1400 ]

• 1

Event counts [ms

• 2

10

• 1

10 1 10 ATLAS Preliminary Simulation

> = 46 µ , <

• e

+

e → Monte Carlo 14 TeV, Z 24 GeV isolated electron trigger Run 2 strategy: < t > = 90.2 ms Run 1 strategy: < t > = 262 ms

Processing time per RoI [ms] 50 100 150 200 250 Normalised Entries

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 1

Two-stage: 1st stage _ _ 0.11 ms ± mean = 23.1 Two-stage: 2nd stage ___ 0.09 ms ± mean = 21.4 Single-stage: . . . 0.34 ms ± mean = 66.2 Tau trigger: Fast Track Finder Data 13 TeV, August 2015, 25 ns running

ATLAS Operations

Andrea Coccaro 16 14 November, 2015 - Trigger Strategies for LLP

SLIDE 17

Tracking Timing Improvements at the HLT

Sometime in the future?

4

10 × Number of spacepoints, 1 2 3 4 5 6 7 Track seeding time [s] 0.5 1 1.5 2 2.5 3 3.5 >=46 µ events < t Monte Carlo 14 TeV t CPU: E5-2695 @ 2.3 GHz GPU: Tesla K80 ATLAS Simulation Andrea Coccaro 17 14 November, 2015 - Trigger Strategies for LLP

SLIDE 18

FTK

Basic logic

◮ parallelisation of the problem by

dividing the detector in η − φ towers

◮ reduction of the data volume by

converting clusters in coarse resolution hits

◮ no patter recognition by comparing

hits to ∼ 109 pre-stored patterns

◮ simplified fit procedure for track

candidates

◮ implemented in FPGA

[GeV]

T

p 10 20 30 40 50 60 70 80 90 Efficiency w.r.t. Offline 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

muon pion

Simulation, no IBL ATLAS η

• 2
• 1

1 2 Efficiency w.r.t. Offline 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

muon pion

Simulation, no IBL ATLAS φ

• 3
• 2
• 1

1 2 3 Efficiency w.r.t. Offline 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

muon pion

Simulation, no IBL ATLAS [mm] d

• 2
• 1.5
• 1
• 0.5

0.5 1 1.5 2 Efficiency w.r.t. Offline 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

muon pion

Simulation, no IBL ATLAS

Winter 2012 Summer 2012 Spring 2013 Summer 2013 2014 2016+ Install Dual Output HOLAs for Vertical Slice Fall Run Vertical Slice in ATLAS Partition, Observer mode Prototypes of all boards, TDR Due ATLAS Approval of TDR Prototype testing Cover barrel for commissioning tests, do full production of AM Boards Install Full FTK System integration tests 2015

Andrea Coccaro 18 14 November, 2015 - Trigger Strategies for LLP

SLIDE 19

Outline

• 1. LLP triggers in calorimeter and muon spectrometer
• 2. LLP triggers in the tracker
• 3. other dedicated triggers
• 4. simplified menu evolution
• 5. exotic approaches (?)

Andrea Coccaro 19 14 November, 2015 - Trigger Strategies for LLP

SLIDE 20

HIP trigger

Specifically designed for monopoles or HIP particles

◮ EM requirement at L1 ◮ no energy requirement in EM2 as in

electron or photon triggers

◮ instead a requirement on the number and

fraction of TRT hits in a narrow region around the L1 seed

Andrea Coccaro 20 14 November, 2015 - Trigger Strategies for LLP

SLIDE 21

Topological selections at L1

Topological triggers at L1 now available in ATLAS

◮ combination of trigger objects from L1 calorimeter and muon systems ◮ can calculate event or signature kinematics or angular relations for making the L1 decision ◮ can actually correlate objects in two adjacent bunch crossings!

• m L1Calo

calculates event kinematics and tral je

min |dPhi(L1_MET, L1 central jets)|

0.5 1 1.5 2 2.5 3

events / 0.07

50 100 150 200 250 300 350 400 MinBias_mu81 mc12 14 TeV ZH125_nunubb mc12 8 TeV Events with at least two L1 central jets

ATLAS Simulation Andrea Coccaro 21 14 November, 2015 - Trigger Strategies for LLP

SLIDE 22

Outline

• 1. LLP triggers in calorimeter and muon spectrometer
• 2. LLP triggers in the tracker
• 3. other dedicated triggers
• 4. simplified menu evolution

[ how much the thresholds changed? ] [ starting information for possibly combining triggers ]

• 5. exotic approaches (?)

Andrea Coccaro 22 14 November, 2015 - Trigger Strategies for LLP

SLIDE 23

Main L1 thresholds at CMS

Unprescaled object pT threshold in 2012, GeV pT threshold in 2015 GeV, (50ns, L=5e33) pT threshold in 2015 GeV, (25ns, L=1.4e34) Single Muon 16 16 20er(*)/25 Double Muon 10+3.5 10+3.5 12+5 Single EGamma 22 25 40 Single Iso EGamma 20 20er(*) 30er(*) Double EGamma 13+7 15+10 22+10 Muon + EGamma 12+7/3.5+12 12+10/5+15 20+10/5+20 Single Jet 128 128 200 Quad Jet 40 40 60 MET 40 50 70 HTT 175 125 175 Double Iso Tau

• 36er(*)

40er(*)

(*) er = eta restricted (restrict to object with large angle wrt beam)

[ H. Brun at LHCP ] Andrea Coccaro 23 14 November, 2015 - Trigger Strategies for LLP

SLIDE 24

Main HLT thresholds at CMS

Unprescaled object pT threshold in 2012, GeV pT threshold in 2015 GeV, (50ns, L=5e33) pT threshold in 2015 GeV, (25ns, L=1.4e34) Single Muon 40 50/45er(*) 50/45er(*) Single Isolated Muon 24 20 24er Double Muon 17+8 17+8 17+8 Single Isolated Electron 27 27 32 Single Photon 150 170 170 Single PF jet 320 450 450 PF MET 120 (80 parked) 170 170 PF HT 750 900 900

work in progress to fine tune some thresholds, 
 taking benefits of the first data

(*) er = eta restricted (restrict to object with large angle wrt beam)

[ H. Brun at LHCP ] Andrea Coccaro 24 14 November, 2015 - Trigger Strategies for LLP

SLIDE 25

Outline

• 1. LLP triggers in calorimeter and muon spectrometer
• 2. LLP triggers in the tracker
• 3. other dedicated triggers
• 4. simplified menu evolution
• 5. exotic approaches (?)

Andrea Coccaro 25 14 November, 2015 - Trigger Strategies for LLP

SLIDE 26

Alternative approaches

Data parking / delayed stream

◮ if computing resources for reconstruction are limited raw data can be saved on tape and

processed later on

◮ already implemented in Run-I and some (prompt) analysis profited

Data scouting / turbo stream

◮ DAQ limitation is bandwidth not rate! ◮ bandwidth = event rate times event size ◮ making the event lighter in size effectively means increasing the acquisition rate ◮ analyses on HLT objects

Some links for nightly thoughts

◮ topological trigger at LHCb - arXiv 1510.00572 ◮ data scouting at Data Science at the LHC workshop - talk

Andrea Coccaro 26 14 November, 2015 - Trigger Strategies for LLP

SLIDE 27

Data scouting

New particle mass Uncovered! Covered Coupling of new particle to quarks Run 1

Challenging idea

◮ only HLT jets being saved ◮ probe low-coupling low-mass di-jet resonances

Andrea Coccaro 27 14 November, 2015 - Trigger Strategies for LLP

SLIDE 28

Conclusions

Trigger strategies are evolving fast

◮ new ideas, new hardware, better algorithm, larger bandwidth, menu fine tuning ◮ not everything covered in this talk and not much public material to show

Points of attention

◮ better coverage for displaced decays of neutral particles in the tracker ◮ collimated topologies

◮ boosted lepton-jets ◮ photon-jets

◮ more combined triggers

◮ targeting exotic Higgs decays ◮ if needed with a mixture of prompt and displaced objects

◮ don’t forget about the backgrounds

◮ triggers running on empty and unpaired bunches ◮ supporting triggers for control regions Andrea Coccaro 28 14 November, 2015 - Trigger Strategies for LLP