Exoplanets & Exomoons

Improving Earth-like Planet Detection in Radial Velocity Using Deep Learning

By Keith Cowing

Status Report

astro-ph.EP

May 24, 2024

Filed under AI, Alpha Centauri B, astro-ph.EP, astronomy, Deep Learning, Earth-like Planet, exoplanet, habitability, HARPS-N, HD10700, HD128621, tau ceti

Improving Earth-like Planet Detection in Radial Velocity Using Deep Learning — TOP: Same as Fig 6 but for the HARPS-N solar spectral time series with the HD10700 sampling. BOTTOM: Exoplanet detection limits on HD128621 HARPS spectra. We derived the detection limit maps by injecting simulated planetary signals covering periods ranging from 10 to 300 days and semiamplitude ranging from 0.4 to 1.1 m/s, into the spectra. Left: Detection limit map in the period-amplitude domain. The red dots indicate the successful detection by the trained neural network of a signal with a FAP > 0.1%. Middle: Amplitude comparison between the injected and recovered signals in the period amplitude domain. The amplitude difference for most of the detected signals is < 30%. Right: Phase difference between the injected and recovered signals in the period-amplitude domain. The phase difference for most of the detected signals is < 0.06. -- astro-ph.EP

Many novel methods have been proposed to mitigate stellar activity for exoplanet detection as the presence of stellar activity in radial velocity (RV) measurements is the current major limitation.

Unlike traditional methods that model stellar activity in the RV domain, more methods are moving in the direction of disentangling stellar activity at the spectral level.

The goal of this paper is to present a novel convolutional neural network-based algorithm that efficiently models stellar activity signals at the spectral level, enhancing the detection of Earth-like planets. We trained a convolutional neural network to build the correlation between the change in the spectral line profile and the corresponding RV, full width at half maximum (FWHM) and bisector span (BIS) values derived from the classical cross-correlation function. This algorithm has been tested on three intensively observed stars: Alpha Centauri B (HD128621), Tau ceti (HD10700), and the Sun.

By injecting simulated planetary signals at the spectral level, we demonstrate that our machine learning algorithm can achieve, for HD128621 and HD10700, a detection threshold of 0.5 m/s in semi-amplitude for planets with periods ranging from 10 to 300 days. This threshold would correspond to the detection of a ∼4M_⊕ in the habitable zone of those stars. On the HARPS-N solar dataset, our algorithm is even more efficient at mitigating stellar activity signals and can reach a threshold of 0.2 m/s, which would correspond to a 2.2M_⊕ planet on the orbit of the Earth.

To the best of our knowledge, it is the first time that such low detection thresholds are reported for the Sun, but also for other stars, and therefore this highlights the efficiency of our convolutional neural network-based algorithm at mitigating stellar activity in RV measurements.

Yinan Zhao, Xavier Dumusque, Michael Cretignier, Andrew Collier Cameron, David W. Latham, Mercedes López-Morales, Michel Mayor, Alessandro Sozzetti, Rosario Cosentino, Isidro Gómez-Vargas, Francesco Pepe, Stephane Udry

Comments: Accepted for publication in A&A
Subjects: Earth and Planetary Astrophysics (astro-ph.EP); Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG)
Cite as: arXiv:2405.13247 [astro-ph.EP] (or arXiv:2405.13247v1 [astro-ph.EP] for this version)
Submission history
From: Yinan Zhao
[v1] Tue, 21 May 2024 23:28:20 UTC (3,812 KB)
https://arxiv.org/abs/2405.13247

Astrobiology,

Keith Cowing

Explorers Club Fellow, ex-NASA Space Station Payload manager/space biologist, Away Teams, Journalist, Lapsed climber, Synaesthete, Na’Vi-Jedi-Freman-Buddhist-mix, ASL, Devon Island and Everest Base Camp veteran, (he/him) 🖖🏻

Follow on Twitter