Time-resolved X-ray spectra of Proxima Centauri as seen by XMM-Newton

TL;DR

This work delivers time-resolved XUV spectra of Proxima Centauri by reanalyzing archival XMM-Newton data with a novel pile-up correction and an adaptive time-binning strategy, enabling ~300 s cadence across 1–920 Å. It combines multi-temperature APEC plasma modeling with a robust regression framework to extract time-dependent spectral components while quantifying uncertainties, and it extends the spectra into the EUV using scaling laws with explicit caveats about flare-state applicability. The study finds that Prox Cen’s X-ray flux varies by up to ~$ imes$20 on short timescales, with larger relative variability at shorter wavelengths, and demonstrates that cadence and modeling choices can significantly alter inferred average fluxes. It also shows that EUV reconstruction carries large, non-negligible uncertainties due to reliance on empirical scaling relations, underscoring the need for long-term, multi-wavelength monitoring to inform atmospheric models of planets such as Prox Cen b.

Abstract

Stellar soft X-ray ([1, 100] Angstrom) and Extreme Ultraviolet (also EUV, [100, 920] Angstrom; jointly, XUV) radiation affects the evolution and chemistry of exoplanet atmospheres. It is however uncertain to what extent the radiation's short-term variability contributes to these effects. We are interested in what this variability might imply for planets around M dwarf stars, and focus on Proxima Centauri (Prox Cen) for three reasons: it is an active M dwarf with high levels of variability; it hosts a likely terrestrial exoplanet within its habitable zone (HZ) that will be a prime target for future direct imaging; its proximity has led to extensive observations. We set out to produce time-resolved XUV spectra of Prox Cen that will serve as input to atmospheric models, and to characterize the intrinsic variability of the star. We analyzed the entire dataset of archival XMM-Newton observations for Prox Cen. To derive the time-resolved X-ray spectra, we implemented a new pile-up correction, a new adaptive time-binning algorithm, and a time-dependent plasma model selection. The estimated EUV spectrum is based on a published template, that we scale with proposed relationships between X-ray and EUV fluxes. We produced spectra of Prox Cen from 1 to 920 Angstrom over ~260 ks of observations with unprecedented time resolution. The instantaneous X-ray flux of Prox Cen varies between about 20 times and one-fifth of the average value over the available baseline, with significant differences between wavelengths. We further quantify how variability affects the estimated average flux when a limited number of snapshots (each typically of 30 ks exposure) are available, as is common in X-ray surveys. Future investigations of the atmospheres of Prox Cen b should fold in the time variability and uncertainties described here.

Figures (13)

• Figure 1: Upper and lower relative errors on the total integrated X-ray flux as a function of total flux for the 3-temperature model. Vertical dashed lines delimit the by-eye–selected ranges. Red and blue segments represent the average errors in the corresponding ranges.
• Figure 2: Comparison of $\omega_i$ for the different intervals of observation 0049350101. The left axis represents the model, the right axis shows the wavelength-integrated flux light curve ([1, 100] Å). The values of the Akaike weights for each model as a function of time are color coded as indicated by the colorbar.
• Figure 3: XUV spectra obtained using the linear (top panel) and quadratic (lower panel) relations from SF+25, scaled from the minimum, average and maximum X-ray fluxes. The XUV-integrated fluxes are reported on the plot together with the EUV/X-ray ratio for maximum and minimum spectra.
• Figure 4: X-ray flux (1-100 Å; count s$^{-1}$; black-solid line) is plotted for selected snapshot observations, together with the X-ray average over all observations. The optical count rate is plotted as a black dashed line. Colored lines (see the color bar at the top panel) indicate lightcurves in different energy bands, obtained by integrating the spectral fluxes. No axis labels or ticks are displayed for these, as they are plotted only for qualitative comparison. Limits are set to (0.8 x min, 1.2 x max) separately for each band.
• Figure 5: The x-axis shows the flux of the star at 1 AU (integrated from 1 to 920 Å). The flux is logarithmically binned. The right-hand-side y-axis refers to the three bold black and white lines. The three lines share this axis because all quantities are expressed as percentages. The black-dashed line represents the percentage of time Prox Cen is emitting at the corresponding x-axis flux bin. The black continuous line is the cumulative curve of the black dashed line. It is showing the percentage of time the star is emitting above the corresponding x-axis flux value. The white continuous line corresponds to the percentage of the total emitted energy of Prox Cen above the corresponding flux value. The vertical black-dashed lines indicate important relative maxima in the observations, excluding the absolute maximum, which would have been plotted at the extreme right. From right to left, in order from the most to the least intense, they refer to: the secondary peak of the main flare of observation 0049350101 (A), the peak of the main flare of observation 0551120401 (B), the maximum of observation 0801880301 (C), and the peak at the start of observation 0049350101 (D). The red-dashed vertical line is the average flux over all the observations. The left-hand-side y-axis describes the partitioning by wavelength range of each flux bin. They are color-coded by the top color bar. This partitioning is obtained by averaging all spectra in the corresponding flux bin.