kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Unsupervised search for resonant production of semi-visible jets with CMS Run 2 Scouting data
KTH, School of Engineering Sciences (SCI), Physics.
2024 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesisAlternative title
Sökning efter resonant produktion av semi-visible jets med CMS Run 2 Scouting-data med hjälp av unsupervised learning (Swedish)
Abstract [en]

A search for resonant production of semivisible jets, arising from the decay of particles in a QCD-like hidden sector, using CMS Run 2 scouting data is presented. The hidden sector is coupled to the standard model (SM) through a leptophobic Z' mediator which decays to hidden sector particles which subsequently hadronize and decay to visible and invisible particles, forming semivisible jets. The relaxed trigger requirements on the objects in the scouting data stream extend sensitivity to low Z' masses, between 700 GeV and 1500 GeV. The search strategy consists of a cut-based analysis and an unsupervised learning strategy to tag semivisible jets against QCD and ttbar background. The cut-based analysis includes a study of the efficiency of the scouting trigger employed. The trigger efficiency is calculated as a function of the transverse mass M_T, which reaches its turn-on point around 650 GeV with a requirement on the difference in pseudorapidity between the leading jets to be $\Delta \eta$ < 1.5. Additional selections on event variables are applied based on physical constraints and the maximization of signal significance. The unsupervised learning strategy involves utilizing the internal structure of the jets, represented as Lund graphs, and training a graph autoencoder to reconstruct graphs of jets from background processes. The autoencoder is then employed to tag semivisible jets through high reconstruction loss, but the strategy may also be used to tag signals outside the model considered in this analysis, serving as an anomaly detection model. The autoencoder achieves relatively high performance for signals with mediator masses of 1100 GeV and above but has a performance close to that of a random classifier for the lowest mediator masses considered.

Abstract [sv]

En studie av resonant produktion av "semivisible jets" som uppkommer från sönderfallet av partiklar i en QCD-liknande dold sektor med använding av CMS Run 2 scouting-data presenteras. Den dolda sektorn kopplas till standardmodellen genom en leptofob Z'-mediator som sönderfaller till partiklar i den dolda sektorn. Dessa hadroniserar i sin tur och sönderfaller till synliga och osynliga partiklar och bildar semivisible jets. De lösa trigger-kraven på objekt i scouting-datastreamen utökar känsligheten till låga Z'-massor, mellan 700 och 1500 GeV. Sökningsstrategin innefattar en cut-baserad analys och en analys med hjälp av en unsupervised learning-modell för att identifiera semivisible jets mot QCD- och ttbar-bakgrund. Den cut-baserade delen av analysen innefattade en studie av scouting-triggerns effektivitet. Triggereffektiviteten beräknades som en funktion av den transversella massan M_T och når sin påslagspunkt runt 650 GeV, med ett krav på skillnad i pseudorapiditet mellan de två ledande jets på $\Delta \eta$ < 1.5. Ytterligare selektioner på eventvariabler appliceras baserat på fysiska krav och maximering av signalens signifikans. Unsupervised learning-modellen utnyttjar den inre strukturen hos jets, som representeras som Lundgrafer, och en autoencoder tränas att rekonstruera grafer av jets från bakgrundsprocesser. Autoencodern används därefter för att identifiera semivisible jets genom hög rekonstruktionsförlust, men strategin kan även användas för att identifiera signaler utanför den aktuella modellen och användas som anomali-detektionsmodell. Autoencodern uppnår relativt hög prestanda för signaler med mediatormassor från 1100 GeV men dess prestanda för låga massor liknar prestandan hos en sluppmässig klassificeringsmodell.

Place, publisher, year, edition, pages
2024.
Series
TRITA-SCI-GRU ; 2024:328
Keywords [en]
High energy physics, unsupervised learning, semivisible jets, graph autoencoder
Keywords [sv]
Högenergifysik, unsupervised learning, semivisible jets, graf-autoencoder
National Category
Physical Sciences
Identifiers
URN: urn:nbn:se:kth:diva-353690OAI: oai:DiVA.org:kth-353690DiVA, id: diva2:1900086
External cooperation
ETH Zürich
Subject / course
Physics
Educational program
Master of Science - Engineering Physics
Supervisors
Examiners
Available from: 2024-09-23 Created: 2024-09-23 Last updated: 2024-09-23Bibliographically approved

Open Access in DiVA

fulltext(14336 kB)215 downloads
File information
File name FULLTEXT01.pdfFile size 14336 kBChecksum SHA-512
318292d726905ac70803817b183dac6704e270bc67836e289e8b2e66e702eb826ce32eb26834c088fca2a1bfc9c6405b683ed313c5de8f00a7e632990e3148ca
Type fulltextMimetype application/pdf

By organisation
Physics
Physical Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 215 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 151 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf