Name/Affiliation: Stephanie Douglas (Columbia University/American Museum of Natural History)
When good fits go wrong: determining realistic best fits and uncertainties on L dwarf physical parameters
Understanding the true uncertainties on parameters derived from model fits to low- and medium-resolution data will be particularly important with the advent of JWST. We expect NIRSpec on JWST to provide moderate-resolution spectra of hot jupiters and greatly expand the spectroscopic sample of L dwarfs. A typical method for inferring physical parameters (effective temperature, surface gravity, and possibly dust or cloud properties) of brown dwarfs is to fit synthetic spectra to observed data. Synthetic spectra can be found that reproduce the observed data reasonably well, both qualitatively and statistically. However, the parameters implied by these “best-fit” spectra are often unphysical: the good fits have gone wrong. Furthermore, best-fit model parameters are frequently presented in the literature without associated uncertainties, despite knowledge that the models lack important opacity sources and cloud physics. Unlike at low-resolution, fitting synthetic spectra to high-resolution data of M and T dwarfs implies parameters that agree with the current understanding of brown dwarf evolution. Although we can derive more reliable parameters at high resolution, available data is scarce and observations require large telescopes (e.g. NIRSPEC on Keck). In contrast, low- and moderate-resolution data is widely available for L dwarfs (from SpeX, FIRE, Triplespec, etc). Using a sample of normal L dwarfs with low-resolution SpeX data, we are determining realistic uncertainties on model fits to low-resolution data. We are employing Markov chain Monte Carlo methods to robustly characterize degeneracies between model parameters and uncertainties on best-fit values. The overall goal of our project is to compare the parameters implied by fitting synthetic spectra from multiple atmospheric models to observed L dwarf data at multiple resolutions. We will present preliminary results showing how well (or poorly) model fits to low-resolution spectra of L dwarfs can constrain effective temperature, surface gravity, and other physical parameters.