ExoIris: fast exoplanet transmission spectroscopy in Python

TL;DR

ExoIris addresses the challenge of jointly analyzing heterogeneous exoplanet transit data by directly modeling two-dimensional spectrophotometric time series and inferring both wavelength-dependent and wavelength-independent parameters in a single framework. It achieves efficiency and flexibility by decoupling spectral resolution from data resolution with a knot-based radius-ratio interpolation and by leveraging RoadRunner for fast transit evaluations, along with GP-based noise modeling and TLSE/spots handling. The package supports complex joint analyses across instruments and epochs, including transit timing variations, instrumental offsets, and unocculted stellar heterogeneities, enabling robust, self-consistent transmission spectra and atmospheres. The WASP-39 b ERS example demonstrates accurate recovery of JWST spectra and stellar density, illustrating ExoIris’s practical impact for precise atmospheric retrieval and cross-instrument validation in exoplanet characterization.

Abstract

I present ExoIris, a user-friendly Python package for exoplanet transmission and emission spectroscopy. Unlike existing tools, ExoIris models two-dimensional spectrophotometric transit time series directly and supports the joint analysis of multiple datasets obtained with different instruments and at different epochs, as well as modeling stellar spot crossings and the influence of unocculted heterogeneities (the transit light source effect). These features enable a self-consistent estimation of both wavelength-independent and wavelength-dependent parameters. They offer a more robust workflow than the commonly used two-step approach, in which a "white" light curve is fitted first, and the transmission spectrum is then derived from independent fits constrained by the white-light solution. Despite its increased flexibility and robustness, ExoIris remains computationally efficient. A low-resolution transmission spectrum can be estimated from a single JWST NIRISS transit observation in ~5 minutes assuming white noise, and in ~15 minutes when accounting for time-correlated systematics using a Gaussian process noise model, on a standard desktop computer.

Figures (11)

• Figure 1: Example of a spectrophotometric transit model. The first two panels show the power-2 limb-darkening coefficients as splines over wavelength with freely placed knots. The third panel shows the planet's radius ratio modeled in the same way. The rightmost panel displays the model flux as a function of time and wavelength.
• Figure 2: Radius-ratio interpolation schemes supported by ExoIris. The black dots show the posterior median radius-ratio knots with their 68% central posterior intervals, as inferred from the JWST WASP-39 b NIRISS observations presented in Feinstein2023 using five of the supported interpolation schemes (shifted in y for visualization). The black lines show the posterior median transmission spectra, and the colored shading shows their 68% central posterior intervals.
• Figure 3: Example illustrating the use of offset groups to account for instrument- or epoch-dependent additive offsets arising from, for example, errors in bias-level estimation or instrumental systematics. The figure shows four spectrophotometric datasets from two transit epochs, each observed with two detectors. The first column shows the radius ratio as a function of wavelength, the next two columns show the spectrophotometric model for all datasets, and the final two columns display the average light curves. The dataset index is denoted by DS, the offset-group index by BG, and the epoch-group index by EG (see Sect. \ref{['sec:exoiris.ttvs']}). Datasets 0 and 1 were observed at epoch 0 and do not feature an offset between the two detectors. However, datasets 2 and 3, observed at epoch 1, show an offset, with the transit of dataset 3 being, on average, half the transit depth of dataset 1.
• Figure 4: Example illustrating the use of epoch groups (EG) to account for transit timing variations in a joint transmission spectrum analysis with ExoIris. The example includes four spectrophotometric datasets: two from the same epoch observed with different instruments, and two from separate epochs observed with a third instrument.
• Figure 5: Model used for stellar spot crossing events. The temporal shape of each event is modeled as a generalized Gaussian, while the wavelength dependence of the spot amplitude is computed using BT-Settl model spectra by Allard2012a.