Skip to content

hpparvi/ExoIris

Repository files navigation

ExoIris: Transmission Spectroscopy Made Easy

Docs Python package Contributor Covenant Licence PyPI version

ExoIris is a user-friendly Python package designed to simplify and accelerate the analysis of transmission spectroscopy data for exoplanets. The package can estimate a self-consistent medium-resolution transmission spectrum with uncertainties from JWST NIRISS data in minutes, even when using a Gaussian Process-based noise model.

Documentation

Read the docs at exoiris.readthedocs.io.

Key Features

  • Fast modelling of spectroscopic transit time series: ExoIris uses PyTransit's advanced TSModel transit model that is specially tailored for fast and efficient modelling of spectroscopic transit (or eclipse) time series.
  • Flexible handling of limb darkening: The stellar limb darkening can be modelled freely either by any of the standard limb darkening laws (quadratic, power-2, non-linear, etc.), by numerical stellar intensity profiles obtained directly from stellar atmosphere models, or by an arbitrary ser-defined radially symmetric function.
  • Handling of Correlated noise: The noise model can be chosen between white or time-correlated noise, where the time-correlated noise is modelled as a Gaussian process.
  • Model saving and loading: Seamless model saving and loading allows one to create a high-resolution analysis starting from a saved low-resolution analysis.
  • Full control of resolution: ExoIris represents the transmission spectrum as a cubic spline, with complete flexibility to set and modify the number and placement of spline knots, allowing variable resolution throughout the analysis.

Details

ExoIris uses PyTransit's TSModel, a transit model that is specially optimised for transmission spectroscopy and allows for simultaneous modelling of hundreds to thousands of spectroscopic light curves 20-30 times faster than when using standard transit models not explicitly designed for transmission spectroscopy.

A complete posterior solution for a low-resolution transmission spectrum with a data resolution of R=100 takes 3-5 minutes to estimate assuming white noise, or 5-15 minutes if using a Gaussian process-based likelihood model powered by the celerite2 package. A high-resolution spectrum of the JWST NIRISS WASP-39 b observations by Feinstein et al. (2023) with ~3800 spectroscopic light curves (as shown above) takes about 1.5 hours to optimise and sample on a three-year-old AMD Ryzen 7 5800X with eight cores.


© 2024 Hannu Parviainen