A Comparative L-dwarf Sample Exploring the Interplay Between Atmospheric Assumptions and Data Properties

TL;DR

This study compares atmospheric retrievals of five L dwarfs to understand how metallicity and data quality influence atmospheric models, revealing that high SNR data limits are dominated by model uncertainties and highlighting methods to resolve cloud model degeneracies.

Contribution

It presents the first atmospheric retrievals for four L dwarfs, analyzing the effects of metallicity and SNR on atmospheric parameters, and discusses strategies to resolve cloud model degeneracies.

Findings

Similar Teff sources have comparable P-T profiles with metallicity affecting profile offsets.

At SNR ≥80, model uncertainties dominate over measurement errors.

SNR influences confidence in parameters but not the ability to distinguish cloud models.

Abstract

Comparisons of atmospheric retrievals can reveal powerful insights on the strengths and limitations of our data and modeling tools. In this paper, we examine a sample of 5 similar effective temperature (Teff) or spectral type L dwarfs to compare their pressure-temperature (P-T) profiles. Additionally, we explore the impact of an object's metallicity and the observations' signal-to-noise (SNR) on the parameters we can retrieve. We present the first atmospheric retrievals: 2MASS J15261405$+$2043414, 2MASS J05395200$-$0059019, 2MASS J15394189$-$0520428, and GD 165B increasing the small but growing number of L-dwarfs retrieved. When compared to atmospheric retrievals of SDSS J141624.08+134826.7, a low-metallicity d/sdL7 primary in a wide L+T binary, we find similar Teff sources have similar P-T profiles with metallicity differences impacting the relative offset between their P-T profiles in the photosphere. We also find that for near-infrared spectra, when the SNR is $\gtrsim80$ we are in a regime where model uncertainties dominate over data measurement uncertainties. As such, SNR does not play a role in the retrieval's ability to distinguish between a cloud-free and cloudless model, but may impact the confidence of the retrieved parameters. Lastly, we also discuss how to break cloud model degeneracies and the impact of extraneous gases in a retrieval model.