Analysis of GBNCC Survey Data for Radio Pulsar Candidate Detection:
Distinguishing Pulsar, RFI, and Noise Signals
Pulsars are fast-spinning neutron stars, star remnants of supernova explosions composed almost entirely of neutrons.
As the most accurate natural “clocks” in the universe, with extremely consistent spin periods, they are used to test physics theories, as well as to detect cosmic phenomena.
Pulsar search is a difficult, yet important, task with great potential yields for the advancement of astrophysics.
This study investigated different features and criteria to distinguish pulsar, RFI, and noise plots from 127,000 pulsar candidates collected through the Green Bank North Celestial Cap Survey with the Green Bank Telescope in West Virginia.
While evaluating various algorithms, a subset of the SR (search results: manually labeled plots) dataset was used.
This comprised all the pulsar plots (of which there are 385), as well as the equivalent number of the top-rated RFI and noise plots (385 each), for a total of 1155 plots, for the purpose of having equal sample sizes during comparison of the algorithms.
The most efficacious method is the Max-Ratio algorithm applied to the Phase-Frequency subplot, with 70% of the first N (estimated number of pulsars) plots being strong candidates; it was rerun with all 103,000 plots in the SR dataset for a larger sample and with 90%, 70%, 60%, and 50% as the threshold value (in addition to the original 80%).
Additionally, the results were compared with the ATNF pulsar catalog. The results demonstrate that the Max-Ratio algorithm is an effective filter, greatly reducing the data amount for manual review.
Yu-Ting Chang is a sophomore at Henry M. Gunn High School in California. He has been a member of the Pulsar Search Collaboratory, a collaboration between the Green Bank Observatory and West Virginia University (funded by NSF Award #1516512), since fall 2019. Aside from astrophysics research, he enjoys programming, volunteering, and anything Star Wars.