Scholarlink

With new high-precision instruments like ESPRESSO, or the James-Webb Space Telescope (JWST), the bottleneck in exoplanet discovery and characterization efforts is set by magnetic effects of the host stars. Our goal is to mitigate this noise to 10 cm s \(^{-1}\) in radial velocity (RV), and 10 parts-per-millions (ppms) in photometry observations. To achieve this, we use our state-of-the-art StarSim modelling tool to feed deep neural networks. In our first approach using RVs and spectroscopic activity index time series, we were able to reduce the stellar RV root-mean-square (rms) jitter by a factor of 50 for our synthetic data and by a factor of 10 for test cases of real observations. Building on those promising first results, we are now improving and calibrating StarSim using newly acquired observational data, working on the extraction of more sensitive activity tracers and applying our approach to other techniques such as transmission spectroscopy. The results of our work will be crucial for the success of endeavors such as the JWST, or Ariel.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reducing Stellar Noise in Exoplanet Observables Using Machine Learning and StarSim

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Manuel Perger,
  • Guillem Anglada-Escudé,
  • Jordi Blanco-Pozo,
  • Òscar Porqueras-León,
  • Juan Carlos Morales,
  • Ignasi Ribas

摘要

With new high-precision instruments like ESPRESSO, or the James-Webb Space Telescope (JWST), the bottleneck in exoplanet discovery and characterization efforts is set by magnetic effects of the host stars. Our goal is to mitigate this noise to 10 cm s \(^{-1}\) in radial velocity (RV), and 10 parts-per-millions (ppms) in photometry observations. To achieve this, we use our state-of-the-art StarSim modelling tool to feed deep neural networks. In our first approach using RVs and spectroscopic activity index time series, we were able to reduce the stellar RV root-mean-square (rms) jitter by a factor of 50 for our synthetic data and by a factor of 10 for test cases of real observations. Building on those promising first results, we are now improving and calibrating StarSim using newly acquired observational data, working on the extraction of more sensitive activity tracers and applying our approach to other techniques such as transmission spectroscopy. The results of our work will be crucial for the success of endeavors such as the JWST, or Ariel.