Jet tagging with ATLAS for discoveries in Run II

Jet tagging with ATLAS
for discoveries in Run II
Ayana Arce
(Duke University)
November 5th 2014
The Large Hadron Collider
ECM: 7.0 – 8.0
mH = 125.4 +/- 0.4
mH = 125.0 +/- 0.3 (CMS)
ECM: 13 – 14
Discoveries at the LHC
dark matter/low EWSB scale  new physics
The LHC is prepared to find:
top partners
superpartners (squark/gluino)
new gauge couplings
extra dimensions
Inner tracker
|η|< 2.5
σ(pT)/pT = 0.05% pT/GeV + 1%
EM Barrel/Endcap
|η|< 3.2
σ(E)/E = 10%/√(E/GeV)
Forward Calorimeter
3.0 < |η|< 4.9
σ(E)/E = 100%/√E
the ATLAS detector
0 < |η|< 1.7
1.7 < |η|< 3.2
σ(E)/E = 50%/√E
What’s left to discover in
Run II?
Why jets? Why now?
Jet substructure tagging
experimental challenges +
Run I constraints on
models of new physics
using substructure tags
Prospects for discoveries
in Run II
Probing the electroweak scale in
Run I
Probing electroweak symmetry
breaking in Run I
Beyond the electroweak scale
Lessons from Run I
Higgs mass requires us to study a variety of
• large branching fraction to bb
• Searches for exotic di-higgs, etc. require fermionic decay
Next searches must probe multi-TeV mass
• large pT for final-state particles in decay
• parton luminosity requires large acceptance in searches
Hadronic decays and boosted object
Looking towards Run II
We will probe higher
masses/boosts at the
same luminosity…
Jet substructure at ATLAS
Hadronic measurements at ATLAS
EM Barrel/Endcap
|η|< 3.2 : δφ ~ 0.025-0.1
σ(E)/E = 10%/√(E/GeV)
0 < |η|< 1.7: : δφ ~ 0.1
1.7 < |η|< 3.2 : δφ ~ 0.1
σ(E)/E = 50%/√E
Hadronic reconstruction
perturbative shower suggests iterative, pairwise
merging algorithms:
jet reconstruction
Jet reconstruction
stable hadrons
Calorimeter jet
Calorimeter cells
Truth jet
iteratively combine closest pairs of particles
distance = min(pTk) (ΔR/Rmax)
topological clusters
Jet constituent observable
moments: calculations
jet mass
average jet charge
m2 = (Σ Ei )2 – (Σ p i)2
jet functions
from fragmentation functions
jet charge
Boost2012 Report, EPJC 74 (2014)
Krohn et. al. “Jet Charge at the LHC” (2012)
Jet constituent observables: parton
top jet mass
Jet constituent calibration
Cluster constituents calibrated to local
hadron scale
Substructure moments re-calibrated at jet level
Interesting particles are color singlet
Color singlet
Color octet
Charge conservation is powerful
LHC backgrounds are … gluey
q/g tagger
Sensitive variables
Gallicchio and Schwartz, PRL107 (2011)
Color factor (g=3 vs. q=9/4) in
substructure moments leads to
many sensitive variables
High pT BG are mostly light partons
Butterworth, Cox, Forshaw (2002)
top/W tagging variables
n-subjettiness ~ 0
typically combined in a “tag”
Thaler, Van Tilburg (2011)
Splitting scale ~(m/2)2
Top-tagging performance
W-tagging correlations
W-tagging performance
Challenges in
substructure tagging
the LHC environment
Jet grooming
Modeling substructure variables
Theory typically
predicts moments –
tagging uses
Parton showers may
disagree, and require
Modeling substructure variables
artist: M. Swiatlowski
Data-driven efficiency: q/g tag
construct width and ntrk distributions
expected for pure samples
• bin in jet pT, η; fix flavor ratios to
MC predictions
also fix heavy flavor templates (shape
and normalization)
Solve to extract pure templates
Data-driven efficiency: jet
Opposite to leptonic W
Color singlet
Charge bias also
possible in W+jets,
Jet charge validation
W → qq candidate charge
Performance of a W+ tagger
Jet pull validation
W-tagging validation
top-tagging validation
Challenging the SM with
substructure tags
Search for W’  tb in hadronic
Consider new gauge interactions in models
preferring quark/3rd gen couplings
Top tagging variables
small differences in
signal distribution for
WL, WR due to top
Limits on W’
Search for W’  WZ, G*  ZZ in
leptonic Z+jet channel
apply three signal regions (2 jet and 1
Boosted channel backgrounds
Confronting Run II challenges
Strategy for 2015
Tagger calibrations:
• W, top tags: In-situ efficiency/fake
rate measurements from Run I
(being completed)
• better q/g purified samples
• grooming and area
subtraction perform well
• also: track-based pileup
constraints (subjet JVT)
Beyond Run II:
Looking ahead
No evidence of physics beyond the SM in Run I
…but a great laboratory for careful validation of
jet tagging observables in data!
Will hadronic final states show us new physics
first in Run II?