The Clear Sky Corridor: Insights Towards Aerosol Formation in Exoplanets Using An AI-based Survey of Exoplanet Atmospheres

TL;DR

This work presents an AI-enabled framework (Eureka!) to autonomously optimize the reduction and analysis of HST/WFC3 transit data, producing homogeneous transmission spectra across a sizable exoplanet sample. By implementing a fitness-driven parametric optimization, the study demonstrates notable improvements in light-curve quality and spectral extraction, enabling a large-scale comparative exoplanetology analysis. The authors identify a Clear Sky Corridor in mass–temperature space where water-band features are enhanced, validate temperature-dependent aerosol formation trends for both Jovians and sub-Neptunes, and link these findings to metallicity and cloud/haze formation. The results offer a scalable pathway for upcoming JWST studies, promising more rapid and reliable atmospheric characterization across a broader wavelength range and chemical inventory, thereby advancing our understanding of exoplanet atmospheres.

Abstract

Producing optimized and accurate transmission spectra of exoplanets from telescope data has traditionally been a manual and labor-intensive procedure. Here we present the results of the first attempt to improve and standardize this procedure using artificial intelligence (AI) based processing of light curves and spectroscopic data from transiting exoplanets observed with the Hubble Space Telescope's (HST) Wide Field Camera 3 (WFC3) instrument. We implement an AI-based parameter optimizer that autonomously operates the Eureka pipeline to produce homogeneous transmission spectra of publicly available HST WFC3 datasets, spanning exoplanet types from hot Jupiters to sub-Neptunes. Surveying 42 exoplanets with temperatures between 280 and 2580 Kelvin, we confirm modeled relationships between the amplitude of the water band at 1.4um in hot Jupiters and their equilibrium temperatures. We also identify a similar, novel trend in Neptune/sub-Neptune atmospheres, but shifted to cooler temperatures. Excitingly, a planet mass versus equilibrium temperature diagram reveals a "Clear Sky Corridor," where planets between 700 and 1700 Kelvin (depending on the mass) show stronger 1.4um H2O band measurements. This novel trend points to metallicity as a potentially important driver of aerosol formation. As we unveil and include these new discoveries into our understanding of aerosol formation, we enter a thrilling future for the study of exoplanet atmospheres. With HST sculpting this foundational understanding for aerosol formation in various exoplanet types, ranging from Jupiters to sub-Neptunes, we present a compelling platform for the James Webb Space Telescope (JWST) to discover similar atmospheric trends for more planets across a broader wavelength range.

Figures (6)

• Figure 1: A visualization of local minimum vs. global minimum. In AI-based optimization, the solution should converge to a minimum by design. The type of AI used for optimization will determine which type of minimum is achievable. Global minima are the best-solutions in the entire search-space. Local minima can be, but are not necessarily the best solutions possible Charbonneau1995Rahmat1999. In this example, a search space for two simultaneously-solved variables is shown in the X and Y axes. The fitness score, shown on the Z-axis, is based on application-specific criteria. Here, lower fitness scores are more optimal solutions.
• Figure 2: Top - An illustration of how each individual is "sequenced" with an assortment of variable-value pairs relative to transit processing. Bottom - The process for how optimization via parametric sweeps and genetic algorithms operate. While parametric sweeps only optimize to local minima, their optimization is very repeatable, and thus are preferred to genetic optimizers for this application.
• Figure 3: An example output from Eureka$!$'s HST optimizer, demonstrating sequential optimization of the pipeline's parameters. For simplicity, the optimization shown here selects the best parameter values solely based on the MAD of the white light curve generated (MAD$_{WLC}$). As each parameter's optimal value is selected, it is stored and used in the optimization of the next parameter, yielding the best sequentially-optimized solution.
• Figure 4: Two separate case studies demonstrating proof-of-concept for the proposed AI-based processing, with HST WFC3 observations. Top - White light and 2D light curves after initial processing and after AI-based processing. Bottom - In this scenario, Eureka$!$'s optimal estimates for the spectrum location were very poor, yielding an unusable data reduction. The AI corrected for this during its optimization, producing an outstandingly-improved result.
• Figure 5: Transmission spectra of the 20 exoplanets analyzed as part of this survey.