Razvoj algoritma za identifikaciju hadronskih mlazova nastalih raspadom ultrarelativističkih Higgsovih bozona u dva b kvarka korištenjem nadograđenog CMS piksel detektora

master's thesis

• Title (english): Development of an algorithm for identification of boosted Higgs bosons decaying in two b quarks using the upgraded CMS pixel detector
• Author: Matej Roguljić
• Mentor: Dinko Ferenček (mentor)
• Mentor: Damir Bosnar (mentor)
• Committee member: Damir Bosnar (predsjednik povjerenstva)
• Committee member: Dinko Ferenček (član povjerenstva)
• Committee member: Krešimir Kumerički (član povjerenstva)
• Committee member: Mirko Planinić (član povjerenstva)
• Committee member: Ivica Smolić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Physics); Zagreb
• Defense date and country: 2017-07-14, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Physics

Abstract

U prvoj polovici 2017. godine izvršena je nadogradnja CMS piksel detektora sa Phase 0 na Phase 1 konfiguraciju. U ovom diplomskom radu ispitan je učinak nadogradnje piksel detektora na algoritam za dvostruko b označavanje. Učinak je određen usporedbom performansi algoritma na simuliranim podacima prije i nakon nadogradnje i rezultati pokazuju da se performanse algoritma poboljšaju korištenjem Phase 1 detektora. Također je izvršen i novi trening algoritma na Phase 1 podacima čime je dobiveno dodatno poboljšanje performansi. Usporedbom Phase 1 i Phase 0 treninga na odgovarajućim podacima u rasponu transverzalnog impulsa hadronskih mlazova od 300 do 1000 GeV pokazano je da Phase 1 algoritam ima bolje performanse na svim radnim točkama. Relativno poboljšanje iznosi oko 10% na radnim točkama visoke čistoće signala što se polako smanjuje smanjenjem čistoće. Utvrđeno je da poboljšanju najviše doprinosi bolja mogućnost raspoznavanja b hadronskih mlazova od hadronskih mlazova nastalih iz lakih kvarkova i gluona. Ispitana je ovisnost algoritma za b označavanje o transverzalnom impulsu hadronskog mlaza i značajne se degradacije performansi pojavljuju tek od 1000 GeV. Dodatno poboljšanje algoritma dobiveno je korištenjem bolje konfiguracije parametara metode stabla odlučivanja koja se koristi tijekom treniranja.

Abstract (english)

The CMS pixel detector was upgraded from Phase 0 to Phase 1 configuration in the first half of 2017. The effect of the upgrade on the double b tagging algorithm is studied in this master’s thesis. It is determined by comparing the performance of the algorithm on the simulated data before and after the upgrade. Results show that the Phase 1 upgrade improves the performance of the double b tagging algorithm. Retraining of the algorithm using Phase 1 data is also performed to get additional improvement. Comparison of Phase 1 and Phase 0 training in the 300–1000 GeV transverse momentum range shows that Phase 1 training performs better on the whole range of working points. Relative improvement starts at about 10% for the high signal purity working points and slowly decreases with decreasing signal purity. It is determined that the upgrade mostly improves discrimination between b hadron jets and light-flavour jets. The stability of the algorithm with respect to the jet transverse momentum is confirmed and significant degradation of performance starts only above 1000 GeV. Further improvement of the algorithm is achieved by using a more optimal set of parameters in the training of the boosted decision tree

• Keywords: CMS piksel detektor; Higgsov bozon; ultrarelativistički režim; b označavanje; dvostruko b označavanje
• Keywords (english): CMS pixel detector; Higgs boson; ultrarelativistic limit; b tagging; double b tagging
• Language: croatian
• URN:NBN: urn:nbn:hr:217:254326
• Study programme: Title: Physics; specializations in: Research; Course: Research; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: magistar/magistra fizike (magistar/magistra fizike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2017-08-01 09:33:30