Trigger Level Analyses in ATLAS (slides written in collaboration - - PowerPoint PPT Presentation

Trigger Level Analyses in ATLAS (slides written in collaboration Lund/OSU)

William Kalderon
 (Lund University) Antonio Boveia 
 (Ohio State University) Caterina Doglioni (Lund University) Oxana Smirnova (Lund University) Eric Corrigan (Lund University) Caterina Marcon (Lund University) Emma Tolley (Ohio State University)

+ more Lund ATLAS students/seniors/engineers

2

Trigger systems in ATLAS/CMS

ATLAS CMS

Level-1: 
 custom hardware Software HLT:
 20k cores 1 kHz to storage 40 MHz bunch crossing

• H. Brun, LP 2015

40

3

Limitations to recording all data

Bandwidth = Event rate x Event size

Limited by:
 fast read-out of o(100M) detector channels computing resources (reconstruction) disk storage (saving for further processing)
 everyone else’s favourite physics channel LHC: 40 MHz 
 ATLAS: 1 kHz
 LHCb: 12.5 kHz CMS: 1 kHz (Reconstructed) ATLAS: o(MB) LHCb: ~100 kB CMS: o(MB)

Probing for low-rate processes is important: LHC luminosity will increase but energy will not. Readout bandwidth is an important limitation of searches when irreducible backgrounds are large.

4

Dark Matter mediators constraints and DM relic density

arXiv:1503.05916

SM SM Med. DM DM SM SM Med. SM SM

gq g𝜓 gq gq

Mediator mass [GeV] Dark Matter mass [GeV]

68% 27% 5%

Ordinary Matter Dark Matter Dark Energy

(quark → jet) (quark → jet) (invisible) (invisible)

Dobrescu, Yu Phys Rev D 88 035021 (2013)

5

Dijet Resonances: Constraints on Coupling Values vs. Mass

Coupling of new particle to quarks

= 6x gq

6

Dirac WIMP mediators: s- and t-channel vector/axial-vector/scalar/pseudo-scalar MET+heavy flavor, W, Z, and Higgs

Mediator Mass [TeV] 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 DM Mass [TeV] 0.2 0.4 0.6 0.8 1 1.2

DM Simplified Model Exclusions April 2016 Preliminary ATLAS

= 1

DM

= 0.25, g

q

g

Axial-vector mediator, Dirac DM

Perturbative unitarity D M M a s s = M e d i a t

• r

M a s s × 2 = 0.12

2

h

c

Ω Thermal relic = . 1 2

2

h

c

Ω T h e r m a l r e l i c

< 0.12

2

h

c

Ω

• Phys. Rev. D. 91 052007 (2015)

Dijet 8 TeV

• Phys. Lett. B 754 302-322 (2016)

D i j e t 1 3 T e V

arXiv:1604.07773

+jet 13 TeV

miss T

E

arXiv:1604.01306

13 TeV γ +

miss T

E

7

8

Trigger-Level Analysis technique

Record only necessary information for jet search: jets Use information already available to make the decision: HLT jets unprescaled prescaled Event size reduced to 5% of fully recorded event Reduced size → increase unprescaled trigger rate

9

Statistics increase from Trigger-Level Analysis technique

Mass of di-jet system
 (DM mediator mass) Number of events

Fully 
 recorded events

Partially 
 recorded
 (smaller size) events

Mass of di-jet system
 (DM mediator mass) Number of events

Prescaled events

Signal QCD background Data

Trigger-Level Analysis Standard analysis

10

Statistics increase from Trigger-Level Analysis technique: dijet invariant mass

Statistics increase Performance comparison:


trigger/offline response = unity this requires custom offline calibrations for trigger jets

11

Results

Dijet + ISR technique

12

ATLAS MET+X and Dijet Searches

Benchmark model points: LHC DM WG, arXiv:1703.05703

13

Outlook

TLA technique powerful when:

• use a subset of the detector information for the search
• object already reconstructed in the trigger and close enough to offline

Other detectors that could help TLAs:

• pile-up subtraction
• rely on information computed in trigger and only write out summary information (real-

time analysis) TLA for SMARTHEP PhD project purposes:

• combine background reduction and TLA: 


dijet+ISR TLA

• move towards real-time trigger calibrations for LHC upgrade

Benchmark model: 
 axial vector DM mediator

SM SM Med. SM SM

SM

gq gq

ATLAS dijet+ISR photon CDF dijet UA2 dijet Projection of CMS fat-jet+ISR jet

ATLAS dijet TLA

ATLAS high-mass dijet

m [GeV]

Z'

200 300 1000 2000 gq 0.05 0.1 0.15 0.2 0.25 0.3 0.35

ATLAS dijet+ISR jet ATLAS dijet+ISR jet TLA ATLAS dijet+ISR photon TLA

Additional Slides

15

Trigger Level Analysis: search

Jet pT > 185, 85 GeV |y*| = |y1-y2|/2 < 0.6

(rejects forward-peaking t-channel QCD processes)

mjj > 460 GeV |y*| = |y1-y2|/2 < 0.3

(reaches lower in mass due to forcing more central, higher pT jets)

mjj > 396 GeV Data-driven background fit (UA2 fcn) Most discrepant region (p-value 0.44) ATLAS-CONF-2016-030 Only probing below M ~ 1.2 TeV (high-mass analysis takes over)

16

Jet performance for TLA

Performance of trigger jets comparable to that of offline jets Improvements benefit jet trigger as a whole

17

Trigger Level Analysis: results

ATLAS-CONF-2016-030 Limits on Gaussian-shape resonances (for reinterpretation) Constraints on DM mediator couplings

Introduction Triggering outside the box Differently timed analyses Dark Matter and TLA

18

Data parking / delayed stream

18

Bandwidth = Event rate x Event size

If computing resources for reconstruction limited: park the raw data and wait (delay) until everything else is processed

Extra bandwidth = Event rate x Event size processed later

Run 1

Introduction Triggering outside the box Differently timed analyses Dark Matter and TLA

19

ATLAS delayed stream results

19

Number of events Mass of di-jet system


Actual events recorded with standard trigger Signal Actual recorded events from delayed triggers

Other analyses using delayed stream in ATLAS/CMS:
 SUSY search for RPV stops
 Dijet angular analysis… Higgs → bbar Fully hadronic top DM searches… lphaT Run 1

arXiv:1407.1376