Optimizing Higgs Analysis at DØ
John Sandy – SULI Intern
(Texas Tech University)
Mentors: Michael P Cooke and Ryuji Yamada
(Fermilab National Accelerator
Laboratory)
John Sandy - 8/9/12 1
Optimizing Higgs Analysis at DØ
Physics – The Standard Model and the Higgs Mechanism
The Standard Model of Particle Physics:
• Three generations of matter• Quarks and Leptons• Force Carriers
Initially predicted that particles should be massless!
Peter Higgs and his colleagues (Englert, Brout, Guralnik, Hagen, Kibble) came to rescue in 1964 and introduced what is now known as the Higgs Mechanism.
John Sandy - 8/9/12 2
Optimizing Higgs Analysis at DØ
The Higgs Boson
• Excitation of the Higgs field (analogous to photon in E&M field)
• Gives mass to all particles
• Interacts more frequently with more massive particles, attributing more mass to them
The Higgs Field Two particles attempting to move across the Higgs Field.
More massive
Less massive
John Sandy - 8/9/12 3
Optimizing Higgs Analysis at DØ
Higgs Boson Signal vs. Background
The Higgs boson has been notoriously difficult to search for:
• Complex, mass dependant branching ratios
• Low production rates
• Lots of SM backgrounds
John Sandy - 8/9/12 4
Optimizing Higgs Analysis at DØ
The DØ Detector
• Inner silicon detector
• Scintillating fiber tracking system
• 2 tesla solenoid
• Uranium liquid argon calorimeter • Large muon system
• 900,000 individual channels
The Tevatron provided proton-antiproton collisions with a center of mass (CM) energy of 1.96 TeV
John Sandy - 8/9/12 5
Optimizing Higgs Analysis at DØ
Event Displays with D0Cafvis
John Sandy - 8/9/12 6
Optimizing Higgs Analysis at DØ
The Higgs Boson Search at DØ
My search is focused on four Higgs final states:
H → WW → lνlν
WH → lνbb
ZH → llbb
ZH → ννbb
Pre-selection:
• Triggering
• Lepton Tagging
• “b-tagging”
Electron / muon events
0tag, 1tag, 2tag, loose, medium, tight
John Sandy - 8/9/12 7
Optimizing Higgs Analysis at DØ
The Higgs Boson Search at DØ
Boosted Decision Trees (BDTs)
Massjj > 120
PassFail
? ?
pp ff
… and so on …
S / B
BDT makes cuts based on signal purity parameters
Never throw out any data, so no chance of losing signal
Bins that pass more cuts have higher signal to background ratio
John Sandy - 8/9/12 8
Optimizing Higgs Analysis at DØ
Using BDTs to train Multi-Variate Analyses (MVAs)
Create a variable list to build a BDT
Run analysis framework and prune variable list:
• Variable Importance
• Variable Correlation
• Variable Modeling ….
I used BDTs to train MVAs for separating Higgs boson signal from the tt, V + jets, and VV backgrounds
John Sandy - 8/9/12 9
Optimizing Higgs Analysis at DØ
Variable modeling based on simulation:
Well modeled Poorly modeled
Using BDTs to train Multi-Variate Analyses (MVAs)
John Sandy - 8/9/12 10
Optimizing Higgs Analysis at DØ
Using BDTs to train Multi-Variate Analyses (MVAs)
BDTs expected signal and background separation
My own training for Higgs vs. VV background discrimination
Multijet training that I am helping train
John Sandy - 8/9/12 11
Optimizing Higgs Analysis at DØ
Final BDTs (Low Mass)After training, the final BDTs are produced and can be used to separate signal from background
John Sandy - 8/9/12 12
Optimizing Higgs Analysis at DØ
Final BDTs (High Mass)
John Sandy - 8/9/12 13
Optimizing Higgs Analysis at DØ
Final BDTs for 2-Tight Tag Events (Low Mass)
John Sandy - 8/9/12 14
Optimizing Higgs Analysis at DØ
Conclusions
The training of secondary MVAs has shown improvement in other channels (4jet)
Further training may improve modeling
When training is fully completed, improved MVAs will improve overall Higgs search at the Tevatron, allowing for stronger results
Recompiled D0Cafvis will allow DØ collaborators to use event displays in future talks and presentations (already been used at DØ collaboration meeting in UK)
John Sandy - 8/9/12 15
On July 4th CERN announced a new “Higgs like” particle using γγ and Z-Z channels
DØ and CDF recently published results claiming evidence for such a particle using the bb channel
Optimizing Higgs Analysis at DØ
This work was supported by the Experimental Particle Physics Division at Fermilab National Accelerator Laboratory, U.S Department of Energy Office of Science’s and Office of Workforce Development for Teachers and Scientists. I would like to thank my mentors Ryuji Yamada and Michael P. Cooke. I would also like to thank Anthony Podkowa and Herb Greenlee for their assistance.
Acknowledgments
John Sandy - 8/9/12 16
Optimizing Higgs Analysis at DØ
Backup
John Sandy - 8/9/12 17
Top Related