Overestimated Pressure Broadening Misleads Model Spectra in Cool M Dwarf Stars

TL;DR

The paper addresses the mismatch between observed TRAPPIST-1 spectra and 1D stellar models, notably the overpredicted FeH Wing-Ford band at $0.99~\mu$m and the water-band continuum shapes. Using the MPS-ATLAS radiative transfer code with 1D PHOENIX temperature-pressure structures, they generate synthetic spectra across a range of van der Waals broadening strengths. They find that the best match to both the FeH feature and the water-band pseudocontinuum occurs with minimized van der Waals broadening, suggesting that Sun-like broadening does not apply to cool M dwarfs. This has practical impact on deriving stellar properties and exoplanet atmospheres around M dwarfs, and motivates future work toward improved pressure-broadening treatments and eventually 3D modeling.

Abstract

Available one-dimensional stellar models fail to reproduce the observed spectrum of the ultracool M dwarf TRAPPIST-1. In particular, current models predict strong iron hydride (FeH) absorption due to the Wing-Ford bands at 0.99$μ$m, yet this spectral feature is only weakly present in TRAPPIST-1 and other mid-to-late M dwarf stars. Additionally, the shape of the continuum between the water bands in the near-infrared does not match between models and observations. Here, we show that assumptions about pressure broadening, specifically van der Waals broadening, have a dramatic effect on modeled broadband spectral features. We use Merged Parallelized Simplified-ATLAS to generate synthetic spectra over a range of van der Waals broadening strengths, adopting 1D PHOENIX temperature-pressure structures. We find that minimal broadening best matches the observed FeH profile at 0.99$μ$m and in the pseudocontinuum between the large water bands. These results suggest that broadening prescriptions derived for Sun-like stars are not valid for lower-mass stars and that pressure broadening for molecular lines in cool stellar atmospheres must be reevaluated. Refining pressure broadening treatments will improve the accuracy of M dwarf spectral models, enabling more reliable determinations of stellar properties and atmospheric compositions of planets orbiting M dwarfs.

Figures (5)

• Figure 1: Spectra of different stellar types. On the left, scaled and vertically shifted flux from NewEra stellar grid is shown on the y-axis and wavelength is shown in microns on the x-axis. On the right, the corresponding pressure-vs-temperature profiles are shown, with temperature in Kelvin on the x-axis and pressure in bars on the y-axis. Prominent atomic and molecular features are annotated Cushing2005Rajpurohit2018. Molecular features increase as we move toward cooler stars. In particular, the FeH and water features (shaded blue bands) stand out in the M8 model and are not present in hotter stars.
• Figure 2: TRAPPIST-1 observations compared with PHOENIX and NewEra stellar grids. In the upper panel, the scaled flux is shown on the y-axis and wavelength is shown in microns on the x-axis. Two PHOENIX spectra are shown for a star of 2600 K with log($g$) of 5.0 (gold) and 3.5 (magenta), respectively. The NewEra grid spectra is shown (dashed blue) for a 2600 K with log($g$) of 5.0 star. NIRISS/SOSS data for TRAPPIST-1 are shown for comparison (black dots). The deviation between each synthetic spectra and the data is shown as a percentage in the lower panel. Disagreement between the synthesized and observed spectra is most pronounced around the FeH feature at $0.99\mu$m and between the large water features. Decreasing log($g$) significantly improves the fit to the FeH feature for the PHOENIX grid. However, a surface gravity of log($g$)=3.5 is consistent with a subgiant star, not an M dwarf. Observations show that TRAPPIST-1 is a main-sequence star with log($g$)$\approx5.2$Agol2021. This discrepancy points to the need to improve current 1D models. With NewEra, we can now better match stellar data with reasonable stellar parameters. However, while NewEra is more consistent with the FeH feature, disagreement persists in the pseudocontinuum between the large water bands centered around 1.4 and 1.9 .
• Figure 3: Comparison of spectra synthesized using a range of van der Waals (vdW) broadening strengths with observed TRAPPIST-1 data. For the upper two panels, the upper y-axis shows the normalized flux, while the x-axis shows the wavelength in microns. Modeled spectra are shown with lines at R$=$200,000 (top) and compared with the data at R=120 (middle). We use 1D PHOENIX atmospheric structure (pressure--temperature profiles) for a star of 2600 K and log($g$)=5.0 as input to MPS-ATLAS. The lowest y-axis shows the percentage deviation between the data and synthesized spectra at R=120. Reducing van der Waals broadening leads to better agreement, though not a complete match between the calculations and observations.
• Figure 4: Effect of van der Waals broadening on individual spectral lines. The normalized flux is shown on the y-axis and the wavelength is shown in microns on the x-axis. The models are shown at R=200,000 and shown within the FeH Wing-Ford band. The lines are colored according to the fractional amount of van der Waals (vdW) broadening with increased broadening shown in lighter colors.
• Figure 5: Comparison of broadening implementations. The normalized flux is shown on the y-axis and wavelength is given in microns on the x-axis. We show the impact of different broadening implementations at high resolution (R=500,000). Measured EXOMOL is shown in purple, Kurucz approximation in dark blue, Unsöld approximation is shown in light blue, and no van der Waals broadening is shown in green. The EXOMOL broadening parameters are the most consistent with using no van der Waals broadening compared with the earlier approximations.