Title: DIAmante TESS Autoregressive Planet Search (DTARPS) Year 2

Abstract: Exoplanet transit searches continually expand to cover larger fractions of the potential planetary host star population, bringing the challenge of analyzing progressively larger datasets without an overwhelming number of false positives. The observed stellar light curves also often have complicated autoregressive noise concealing transits with small depths. We analyze light curves extracted from ~1 million TESS Year 2 northern hemisphere Full Frame Images by the DIAmante project using the Autoregressive Planet Search (ARPS) methodology. ARPS de-trends light curves using autoregressive modeling, detects planetary transits with our Transit Comb Filter, and refines the list of candidates with a sophisticated machine learning classifier (RandomForestSRC). The classifier considers dozens of light curve features, and is trained on injected planetary signals and against injected eclipsing binary signals. We then use the classifier from Year 1 (Melton et al. 2023) which achieves a true positive rate of ~92% with a false positive rate (FPR) of ~0.3%. We are exploring the use of gradient boosting in decision tree classifiers with the XGBoost code; preliminary findings suggest a strong reduction in FPR will emerge. Classification is followed by human vetting and ephemeris matching in order to further reduce contamination. We apply the methodology to 962,663 stars in Year 2 data and find 1,505 candidate exoplanet systems in a preliminary list. The analyses of Year 2 as well as Year 1 data have yielded hundreds of high confidence planet candidates, many of which fill in the “Neptune desert.” We enthusiastically invite Penn State exoplanetary astronomers to join in spectroscopic follow-up of northern DTARPS candidates.

Host: TBA

Seminar held in 538 Davey or please email CEHW-SEMINAR-QUESTIONS@lists.psu.edu to attend virtually.