Machine learning for exoplanet detection in high-contrast spectroscopy - Revealing exoplanets by leveraging hidden molecular signatures in cross-correlated spectra with convolutional neural networks

Aug 5, 2026

Upload by Aria

Context: Next-generation observatories and instruments require robust techniques to detect and characterize faint, close-in exoplanets. Molecular mapping and cross-correlation spectroscopy use molecular templates to separate a planet signal from the host star, but relying on signal-to-noise metrics can miss discoveries due to strong assumptions about noise. Aims: MLCCS improves detection sensitivity using weaker assumptions, focusing on molecular presence. Methods: MLCCS employs perceptrons and 1D convolutional neural networks in the cross-correlated spectral dimension and is tested on mock data using SINFONI K-band noise. Results: Large gains in detectable planet numbers and completeness, plus improved confidence and conspicuity. Conclusions: After training, MLCCS enables sensitive, rapid, adaptable exoplanet and molecular-species detection.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages26

Reading Time1 h 6 min

AccessFree to download

Abstract

Context and Aims

Methods

Results and Conclusions

1. Introduction

Atmosphere characterization goals

Existing approaches and motivations

A&A, 689, A143 (2024) [https://doi.org/10.1051/0004-6361/202449149](https://doi.org/10.1051/0004-6361/202449149)[ ](https://doi.org/10.1051/0004-6361/202449149)© The Authors 2024 &~~t~~ronomy~~t~~rop~~h~~ys~~i~~cs Machine learning for exoplanet detection in high-contrast spectroscopy Revealing exoplanets by leveraging hidden molecular signatures in cross-correlated spectra with convolutional neural networks Emily O. Garvin 1 , 2 , ⋆, Markus J. Bonse 1 , Jean Hayoz 1, Gabriele Cugno3 , Jonas Spiller 1, Polychronis A. Patapis 1 , Dominique Petit dit de la Roche5 , Rakesh Nath-Ranga4 , Olivier Absil4, Nicolai F. Meinshausen2 , and Sascha P. Quanz 1 1 Institute for Particle Physics and Astrophysics, ETH Zürich, Wolfang-Pauli-Strasse 27, 8093 Zürich, Switzerland 2 Seminar für Statistik, ETH Zürich, Raemistrasse 101, 8092 Zürich, Switzerland 3 Department of Astronomy, University of Michigan, Ann Arbor, MI 48109, USA 4 STAR Institute, University of Liège, 19 Allée du Six Août, 4000 Liège, Belgium 5 Département d’Astronomie, Université de Genève, 1290 Versoix, Switzerland Received 2 January 2024 / Accepted 23 June 2024 ABSTRACT Context. The new generation of observatories and instruments (VLT/ERIS, JWST, ELT) motivate the development of robust methods to detect and characterise faint and close-in exoplanets. Molecular mapping and cross-correlation for spectroscopy use molecular templates to isolate a planet’s spectrum from its host star. However, reliance on signal-to-noise ratio metrics can lead to missed discoveries, due to strong assumptions of Gaussian-independent and identically distributed noise. Aims. We introduce machine learning for cross-correlation spectroscopy (MLCCS) . The aim of this method is to leverage weak assumptions on exoplanet characterisation, such as the presence of specific molecules in atmospheres, to improve detection sensitivity for exoplanets. Methods. The MLCCS methods, including a perceptron and unidimensional convolutional neural networks, operate in the crosscorrelated spectral dimension, in which patterns from molecules can be identified. The methods flexibly detect a diversity of planets by taking an agnostic approach towards unknown atmospheric characteristics. The MLCCS approach is implemented to be adaptable for a variety of instruments and modes. We tested this approach on mock datasets of synthetic planets inserted into real noise from SINFONI at the K-band. Results. The results from MLCCS show outstanding improvements. The outcome on a grid of faint synthetic gas giants shows that fora false discovery rate up to 5%, a perceptron can detect about 26 times the amount of planets compared to an S/N metric. This factor increases up to 77 times with convolutional neural networks, with a statistical sensitivity (completeness) shift from 0.7 to 55.5% . In addition, MLCCS methods show a drastic improvement in detection confidence and conspicuity on imaging spectroscopy. Conclusions. Once trained, MLCCS methods offer sensitive and rapid detection of exoplanets and their molecular species in the spectral dimension. They handle systematic noise and challenging seeing conditions, can adapt to many spectroscopic instruments and modes, and are versatile regarding planet characteristics, enabling the identification of various planets in archival and future data. Key words. methods: data analysis – methods: statistical – planets and satellites: atmospheres – planets and satellites: detection 1. Introduction Spectroscopic observations of substellar companions are crucial for advanced characterisation of exoplanet and brown dwarf atmospheres from emission and transmission spectra. The primary objectives in characterising these atmospheres consist of constraining the molecular composition, abundances, clouds, and thermal structure of exoplanet atmospheres (e.g., Line et al. 2016 ; Brogi & Line 2019) . These measurements offer valuable insights into the formation history of

FAQs

Why can traditional signal-to-noise metrics miss exoplanet discoveries in cross-correlation spectroscopy?

Because such metrics rely on strong assumptions about noise being Gaussian, independent, and identically distributed, which can be violated in real data.

What is MLCCS and how does it work?

MLCCS is machine learning for cross-correlation spectroscopy. It uses a perceptron and 1D convolutional neural networks to identify molecular patterns directly in the cross-correlated spectral dimension.

How much improvement does MLCCS provide compared with an S/N metric?

For a false discovery rate up to 5%, a perceptron detects about 26 times more planets, and convolutional neural networks increase this up to 77 times, with completeness shifting from 0.7 to 55.5%.

Aug 5, 2026

Upload by Aria