Preparing an unsupervised massive analysis of SPHERE high contrast data with the PACO algorithm

TL;DR

This paper presents an unsupervised, optimized analysis pipeline using the PACO algorithm for high-contrast imaging data, significantly improving exoplanet detection capabilities in archival SPHERE data.

Contribution

It introduces an improved, unsupervised data analysis approach with custom tools and spectral priors, enhancing detection limits and confidence control in direct imaging surveys.

Findings

Achieved up to tenfold contrast improvement at 0.2-0.5 arcsec.

Detected planets down to 5 MJup at 5 au around stars within 60 parsecs.

Identified two exoplanet candidates requiring follow-up.

Abstract

We aim at searching for exoplanets on the whole ESO/VLT-SPHERE archive with improved and unsupervised data analysis algorithm that could allow to detect massive giant planets at 5 au. To prepare, test and optimize our approach, we gathered a sample of twenty four solar-type stars observed with SPHERE using angular and spectral differential imaging modes. We use PACO, a new generation algorithm recently developed, that has been shown to outperform classical methods. We also improve the SPHERE pre-reduction pipeline, and optimize the outputs of PACO to enhance the detection performance. We develop custom built spectral prior libraries to optimize the detection capability of the ASDI mode for both IRDIS and IFS. Compared to previous works conducted with more classical algorithms than PACO, the contrast limits we derived are more reliable and significantly better, especially at short angular separations where a gain by a factor ten is obtained between 0.2 and 0.5 arcsec. Under good observing conditions, planets down to 5 MJup, orbiting at 5 au could be detected around stars within 60 parsec. We identified two exoplanet candidates that require follow-up to test for common proper motion. In this work, we demonstrated on a small sample the benefits of PACO in terms of achievable contrast and of control of the confidence levels. Besides, we have developed custom tools to take full benefits of this algorithm and to quantity the total error budget on the estimated astrometry and photometry. This work paves the way towards an end-to-end, homogeneous, and unsupervised massive re-reduction of archival direct imaging surveys in the quest of new exoJupiters.