
Radio Transient Classification Algorithms for CHIME/FRB Alexander Josephy Department of Physics McGill University Montréal, Québec A thesis submitted to McGill University in partial fulfillment of the requirements of the degree of Master of Science © Alexander Josephy, February 2018 ABSTRACT Fast Radio Bursts (FRBs) are micro-to-millisecond duration transients of extragalactic origin. Despite FRB rate estimates of thousands per sky per day, only 30 sources have been detected in the now decade since the initial discovery, and the progenitor of FRBs remains a mystery. This will soon change, as a dedicated 24/7 real-time search, expected to detect tens of FRBs every day, is in the early stages of commissioning. The search is hosted by a novel software-driven telescope with a ∼250 square degree field of view and a collecting area of ∼8000 m2. The telescope is located at the Dominion Astrophysical Observatory (DRAO), and was designed for the Canadian Hydrogen Intensity Mapping Experiment (CHIME). This thesis follows the development of components in the later stages of the CHIME/FRB pipeline, where the focus shifts from detection to classification. These developments include: algorithms to group detections over multi-dimensional search spaces and uniquely identify single- pulses; an algorithm to sift candidate events resembling radio-frequency interference (RFI), with preparations for the application of machine learning classifiers; a method for localizing bright bursts, exchanging memory and precomputation for sub-millisecond execution time; and an architecture for the event-based pipeline, designed to scale to an arbitrary number of CPU cores and machines. Finally, a stress test of the event-based pipeline was performed, using simulated events based on the known pulsar population. The pipeline comfortably handles 300,000 fake events spanning 24 hours of fake observation, reaffirming that the FRB search will not suffer the pulsar foreground. ii RÉSUMÉ Les sursauts radio rapides (Fast Radio Bursts, FRB) sont transitoires de durée micro à milliseconde d’origine extragalactique. Malgré des esti- mations de taux de FRB de milliers par ciel par jour, seulement 30 sources ont été détectés au cours de cette décennie depuis la découverte initiale, et les progéniteurs des FRB restent un mystère. Cela va bientôt changer, des recherches dédiés en temps réel 24/7, devraient détecter des dizaines de FRB tous les jours, sont dans les premiers stades de mise en service. La recherche est hébergée par un télescope piloté par logiciel avec un champ de vision d’environ ∼250 degrés carrés et une zone de collection de ∼8000 m2. Le télescope est situé à l’Observatoire Fédéral Radioastrophysique (OFR), et a été conçu pour L’Expérience Canadienne de Cartographie de L’Intensité de l’Hydrogène (Canadian Hydrogen Intensity Mapping Experiment, CHIME). Cette thèse suit le développement des composantes dans les dernières étapes de la CHIME/FRB pipeline, où l’attention est sur la classification. Ces développements comprennent: des algorithmes pour regrouper des détections sur des espaces de recherche multidimensionnels et identifier de façon unique des impulsions uniques; un algorithme pour trier les événements candidats ressemblant aux interférences radio-fréquences, avec des préparations pour l’application de classificateurs d’apprentissage automatique; une méthode pour localiser des éclats lumineux, l’échange de la mémoire et précalcul de temps d’exécution de sous-millisecondes; et une architecture pour le pipeline basé sur les événements, conçu pour évoluer vers un nombre arbitraire de cœurs et de machines. Enfin, un test de contraintes du pipeline basé sur les événements a été réalisé, en utilisant des événements simulés sur la population de pulsars connue. Le pipeline gère confortablement 300000 faux événements couvrant 24 heures de observations simulées. iii TABLE OF CONTENTS Abstract................................... ii Résumé................................... iii Acknowledgements.............................v Contribution Breakdown.......................... vi The CHIME Collaboration......................... vii List of Figures and Tables......................... viii 1 An Introduction to Fast Radio Bursts1 1.1 Dispersion............................. 2 1.2 Detection History......................... 7 1.3 FRB Progenitor Models ..................... 11 2 The CHIME Family 13 2.1 Canadian Hydrogen Intensity Mapping Experiment . 14 2.2 CHIME/FRB........................... 15 2.3 CHIME/Pulsar.......................... 16 2.4 The CHIME Telescope...................... 17 3 Pipeline Overview 23 3.1 L1 — Detecting Pulses...................... 23 3.2 L2/L3 — Characterizing Pulses................. 25 3.3 L4 — Taking Action ....................... 27 4 L1B Algorithms 30 4.1 Grouping Triggers......................... 31 4.2 A Second Chance for RFI Removal............... 35 5 L2/L3 Algorithms 40 5.1 Pipeline Architecture....................... 41 5.2 Beam Buffer............................ 45 5.3 Beam Grouping.......................... 46 5.4 Sky Localization ......................... 49 5.5 DM Maps............................. 55 6 Event Simulation 58 6.1 Pulsar Selection.......................... 59 6.2 Faking Events........................... 63 6.3 Stress Test Results ........................ 64 7 Closing Remarks 67 Bibliography................................ 70 iv ACKNOWLEDGEMENTS First of all, thanks to Vicky, for taking me in as an undergraduate and giving me the opportunity to continue on, for all the patience and encouragement, and for the infectious excitement towards astrophysics that has solidified my own. Thanks to the weekly codathon crew, it was an absolute pleasure to be part of such a driven and talented team. We’ve accomplished a lot and I am excited for what the future holds. Similarly, thanks to the CHIME collaboration and the McGill neutron star group; the exposure to such a wide range of world-class expertise on a weekly basis has been amazing. Thanks to everyone at the DRAO, the level of hospitality has made my time on site very special. Thanks to my friends, old and new, for getting me out of my head, out of my apartment, and out of the city. Thanks to Yasmine, for the translation, for the empathy and positivity, for being the ultimate antidote to feeling overwhelmed. Finally, thanks to my parents, for getting me through university and enthusiastically supporting my degree choices, for imparting an appreciation of art, math, hard work, travel, the outdoors, and so much more, for 25 years of unconditional love. v CONTRIBUTION BREAKDOWN Chapter2 serves as an overview of the CHIME cosmology, FRB, and pulsar projects, including a technical overview of the telescope they share. More than just a source of context, this chapter shines a light on the massive amount of work that has preceded the developments outlined in this thesis. As such, the entire collaboration deserves recognition (the members are listed on the following page). Likewise, the work in this thesis fits into a much larger pipeline, which is described in Chapter3. The primary contributors of code for each ‘level’ of the pipeline are as follows: Kendrick Smith, Dustin Lang, Masoud Refiei Ravandi, Utkarsh Giri, and Alex Josephy for L1 (§3.1); Shriharsh Tendulkar, Emmanuel Fonseca, Paul Scholz, Shiny Brar, Ziggy Pleunis, Chitrang Patel, and Alex Josephy for L2/L3 (§3.2); Davor Cubranic, Michelle Boyce, and Chitrang Patel for L4 (§3.3). The overall efforts were coordinated by Victoria Kaspi (Principal Investigator) and Patrick Boyle (Project Manager). Original work is presented in Chapters4 through6. In particular, §4.1,§4.2,§5.2, §5.3, and Chapter6 contain independently developed material. In contrast, the architecture for the event-driven pipeline (§5.1) was substantially influenced by Shiny Brar, Chitrang Patel, and Davor Cubranic. The fast multi-beam localization method (§5.4), was made possible by groundwork performed by Paul Scholz, who continues to manage and develop the overall localization effort. Likewise, constructing DM map interpolators (§5.5) was done as part of a larger FRB classification effort, which is managed by Emmanuel Fonseca. vi THE CHIME COLLABORATION Mandana AmiriΛb, Kevin BanduraΛγg, Philippe BergerΛc, Mohit Bhardwahaγay, Dick BondΛγc, Michelle Boyceγa, Patrick Boyleγa, Shiny Brarγa, Maya Burhandpurkarz, Jatin Chowdhuryz, Jean-Francois ClicheΛa, Liam Connorz, Miles Cranmerγay, Davor Cubranicγb, Greg Davisz, Paul Demorestγf , Meiling DengΛby, Nolan DenmanΛγcy, Matt DobbsΛγa, Rachel DomagalskiΛc, Mateus FandinoΛby, Emmanuel Fonsecaγa, Bryan Gaenslerγc, Kwinten Van Gassenz, Adam GilbertΛa, Utkarsh Giriγey, Deborah Goodγby, Sidhant GulianiΛby, Mark HalpernΛγb, David HannaΛγa, Adam Hincksz, Gary HinshawΛγb, Caroline HoferΛby, Gilbert HsyuΛay, Alex Josephyγay, Vicky Kaspiγa, Peter Klagesz, Tom LandeckerΛγd, Dustin Langγe, Haoran Liaoz, Erik Madsenz, Kiyo MasuiΛγb, Nikola MilutinovicΛb, Arun Naiduγa, Laura NewburghΛh, Cherry Ngγb, Mike NoltaΛc, Neils Oppermannz, Juan Mena ParraΛay, Chitrang Patelγa, Ue-Li PenΛγc, Jeff PetersonΛi, Tristan Pinsonneault- MarotteΛby, Ziggy Pleunisγay, Scott Ransomγf , Masoud Refiei Ravandiγey, Andre RenardΛγc, Paul Scholzγd, Richard ShawΛb, Seth SiegelΛa, Kris SigurdsonΛγb, Graeme Smecherz, Rick SmegalΛb, Kendrick SmithΛγe, Ingrid Stairsγb, Emilie StorerΛay, Shriharsh Tendulkarγa, Ian TretyakovΛc, Keith VanderlindeΛγc, Don WiebeΛb a : McGill University b : University of British Columbia c : University of Toronto d : Dominion Radio Astrophysical Observatory
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages85 Page
-
File Size-