Separating flare and secondary atmospheric signals with RADYN modeling of near-infrared JWST transmission spectroscopy observations of TRAPPIST-1

TL;DR

This study tackles the challenge of flare contamination in JWST transmission spectroscopy of TRAPPIST-1 by combining empirical fiducial flare spectra with time-dependent RADYN flare models to separate flare and potential atmospheric signals in NIRISS/NIRSpec data. The authors construct fiducial flare spectra across multiple $T_ ext{eff}$ bins, perform empirical and physics-based (RADYN) mitigation, and quantify the detectability of secondary atmospheres via injection tests, finding robust sub-100 ppm residuals and 3σ detection capabilities for CO$_2$-like features around 200 ppm. The RADYN fits imply moderate electron-beam heating with associated XUV/FUV/NUV outputs and provide scaling relations to relate JWST-band flares to high-energy fluxes, informing atmospheric escape and photochemistry assessments for TRAPPIST-1. The work emphasizes the need for simultaneous multiwavelength observations to test model predictions and improve flare mitigation, aiming to enable robust constraints on temperate exoplanet atmospheres around flare-active M dwarfs.

Abstract

Although TRAPPIST-1's temperate planets have the highest transmission signals of any known system, flares contaminate 50-70% of transits at the 1000 ppm level, far above 100 ppm secondary atmospheres. Efforts to mitigate flare contamination and assess impacts on radiation environments are each hampered by a lack of empirical spectral analysis and physics-based modeling. We present spectrotemporal analysis and radiative-hydrodynamic modeling of 5.5 hr of NIRISS and NIRSpec observations of 6 TRAPPIST-1 flares of 2.2-8.7x10^30 erg. Flare lines and continua are characterized using grid searches of RADYN beam-heating models spanning 10$^4\times$ in electron beam parameters. Best-fit models indicate these flares result from moderate-intensity beams with emergent electron fluxes of 10^12 erg s^-1 cm^-2 and energies $\leq$37 keV, although all models over-predict the Paschen jump. These models predict XUV, FUV, and NUV counterparts to the infrared peak fluxes of 8.9-28.9x10^27, 4.3-13.9x10^26, and 3.4-11.4x10^27 erg s^-1, respectively. Scaling the flare rate into the XUV suggests flaring contributes 1.35$_{-0.15}^{+2.0}\times$ quiescence yr$^{-1}$. We bin integrations of similar flare effective temperature to construct fiducial flare spectra from 2000-4500 K in order to develop separate empirical and RADYN-based mitigation pipelines. Both pipelines are applied to all 5.5 hr of R=10 data, resulting in maximum residuals from 1-2.8$μ$m of 100-140 ppm and typical residuals of 54$\pm$14 and 65$\pm$17 ppm for the empirical and RADYN-based pipelines, respectively. Injection testing supports 3$σ$ detection capability for CO2 atmospheres with features of 150-250 ppm, with weak evidence (BF$\approx$3) still obtained at 130 ppm. Our results motivate multi-wavelength observations to improve model fidelity and test high-energy predictions.

Figures (8)

• Figure 1: Top: Wavelength-integrated (0.6--2.83 $\mu$m), transit-normalized flare light curves of our sample. Flare temperatures of integrations used for model validation and mitigation are scaled by color. Bottom left: Fiducial-$T_\mathrm{eff}$ flare spectra are computed from similar-temperature integrations. Best-fitting blackbody curves are shown with dashed gray lines. Bottom right: Quiescent features appear in the residual spectra after subtracting the best-fit blackbody curves.
• Figure 2: Top: Peak flare spectrum of each event overlaid with the best-fit RADYN model, where fits are color-coded by reduced $\chi^2$ value. The full model grid is shown for reference in gray, where each gray model is scaled by the filling factor of the highlighted model. Models qualitatively describe line and 0.82--2.83 $\mu$m continuum emission, but predict a large Paschen jump not present in the observations. Daggers denote NIRSpec flares. Bottom: A close-up of the best-fit RADYN line models for the F4 event compared with the NIRISS data. Although not included in RADYN, P$\gamma$ is shown to alleviate blending concerns for the He$_\mathrm{I}$ IRT.
• Figure 3: Flowchart illustrating the empirical flare correction procedure. We note the steps in the physics-based procedure are similar to those in the empirical procedure.
• Figure 4: Left: Best-fit RADYN model and uncertainty region for the F1 peak spectrum, determined from all models with residuals within 100% of the best-fit value. The XUV emission driving atmospheric escape, FUV--NUV emission driving photochemistry, and TESS wavelengths are highlighted. Prediction capability degrades with distance from the NIR. Middle: RADYN spectral energy distributions enable scaling of the TESS-band flare rate into the XUV, FUV, and NUV. Right: Comparison of flaring and quiescent contributions to the XUV radiation environment, where the quiescent relation is reproduced from Fleming:2020. The flaring XUV activity-age relation is constructed from the present-day XUV flare rate, estimated saturation flare rate relative to present-day Paudel:2018, and estimated $\sim$4 Gyr saturation knee Fleming:2020.
• Figure 5: Top: Residual flare contamination of 60--140 ppm in the $R$=10 average detrended spectrum of all 189 integrations following application of the empirical (left, blue) and RADYN model-based (gold, right) pipelines. Bottom: Detection significance versus transmission signal strength for the detrended spectra, derived from injection-and-recovery tests of Lustig-Yaeger:2019 CO$_2$ atmospheric transmission signals scaled from 12--480 ppm. Estimated Bayes Factors of 3, 12, and 150 represent "weak," "moderate," and "strong" detections Trotta:2008. The detection significance curve using the simple Planck fit approach from Howard:2023 is shown for reference (gray).