JWST COMPASS: Insights into the Systematic Noise Properties of NIRSpec/G395H From a Uniform Reanalysis of Seven Transmission Spectra

TL;DR

The paper investigates the systematic noise properties of NIRSpec/G395H transmission spectra using seven uniform COMPASS targets to understand detector systematics and improve time-series modeling. It presents a principal-component-based systematics model for normalized pixel fluxes that mitigates trace-shape variations and shows benefits for low-groups-per-integration observations, with strong systematics in the $2.8$–$3.5\,\mu$m range. Comparing real- versus predicted spectroscopic errors reveals Pandexo underestimates uncertainties by about $5\%$ (NRS1) and $12\%$ (NRS2), and the authors derive new lower limits on metallicity and opaque pressure levels for each planet, comparing to prior COMPASS results. Co-adding spectra across targets yields no compelling common features, but additional transits can break the metallicity–aerosol degeneracy for most targets, suggesting future JWST allocations could unlock the atmospheric characterization of these worlds.

Abstract

JWST has already observed near-infrared transmission spectra of over a dozen super-Earths and sub-Neptunes. While some observations have allowed astronomers to characterize sub-Neptunes in unprecedented detail, small feature amplitudes and poorly-understood systematics have led to ambiguous results for others. Using the first seven targets from the COMPASS program, which is currently surveying 12 small planet atmospheres using NIRSpec/G395H, we investigate these timeseries systematics. We implement a model that uses the principle components of the normalized pixel fluxes to account for variations in the shape and position of the spectral trace. We find that observations with a smaller number of groups-per-integration benefit most profoundly from the use of this model, and that systematics are particularly strong between 2.8 and 3.5 $μ$m. Despite these systematics, \texttt{pandexo} is a relatively accurate predictor of the precision of the spectra, with real error bars on average 5\% larger in NRS1 and 12\% larger in NRS2 than predicted. We compute new limits on metallicity and opaque pressure level for each target and compare these to previous results from the COMPASS program. Next, we co-add spectra from multiple targets to reduce the effective noise in the combined spectra in hopes of detecting transmission features in common between the targets, but this exercise does not yield compelling evidence any signals. We find that a handful of additional transits are sufficient to break the degeneracy between metallicity and aerosols for the majority of our targets, pointing towards the possibility of unraveling the mysteries of these worlds with future allocations of JWST time.