Serviço local de periodograma em GPU para detecção de trânsitos planetários

• Ano de defesa: 2017
• Autor(a) principal: Basile, Antonio Luiz
• Orientador(a): Valio, Adriana Benetti Marques
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Presbiteriana Mackenzie
• Palavras-chave em Português: periodograma; box fitting; detecção de exoplanetas; astroinformática; computação de alto desempenho
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA; CNPQ::CIENCIAS EXATAS E DA TERRA::ASTRONOMIA
• Link de acesso: http://dspace.mackenzie.br/handle/10899/25795

Understanding other stellar systems is crucial to a better knowledge of the Solar System, as is the study of extrasolar planets orbiting the habitable zone of their host star central to the understanding of the conditions that allowed life to develop on our own planet. Presently there are thousands of confirmed planets, mostly detected by the Kepler satellite as they eclipse their host star. This overload of data urges an automatic data search for planetary transit detection within the stellar light curves. Box-Fitting Least Squares (BLS) is a good candidate for this task due to the intrinsic shape of the transiting light curve. Further improvement is obtained by parallelization of the BLS according to the number of bins. Both the sequential and parallel algorithms were applied to six chosen Kepler planetary systems (Kepler-7, Kepler-418, Kepler-439, Kepler-511, Kepler-807, Kepler-943) and to different light curve lengths. In all cases, speedup increased from 3 to 45 times as the number of bins increased, because the performance of the sequential version degrades with an increase in the number of bins, while remaining mainly constant for the parallel version. For smaller planets with longer orbital periods, a large number of bins is necessary to obtain the correct period detection.