Stellar halo subtraction alternative for accreting companions' characterization with integral field spectroscopy: Analytical and on-sky demonstration on the PDS70, HTLup, and YSES1 systems

TL;DR

The paper tackles biases in stellar halo subtraction for faint, accreting companions in integral field spectroscopy by introducing Legendre polynomial modulation (LPM), a linear, parametric approach that models chromatic halo deformation and protects emission-line spectra. Through analytical toy models, simulations, and on-sky MUSE data for PDS70, HTLup, and YSES1, the authors show that LPM preserves line fluxes and profiles while maintaining detection capability, achieving 5σ detections for PDS70 b and c and correcting accretion-rate estimates (e.g., YSES1 b) by about 30% compared with SGF. Their results demonstrate that LPM provides a spectral-information-preserving alternative to the traditional Savitzky-Golay filtering, with a clear bias-variance trade-off that can be tuned via the polynomial degree $\partial$ and PSF-decomposition insights. The method is flexible, extendable, and well-suited for integration with additional faint-signal search algorithms and for applications beyond emission lines, such as molecular mapping and regularized inverse problems in exoplanet characterization.

Abstract

Context. Medium-resolution IFS, such as MUSE at the VLT, are equipped to detect the emission lines of faint accreting companions when associated with dedicated stellar halo subtraction methods. We recently proposed a new approach based on polynomial modulations of a stellar spectrum estimate across the field of view, with orthogonal polynomials and lines masking. Aims. We seek to highlight and quantify analytically and on real data the benefits of this new approach over the one classically used, particularly with regard to distortions of the extracted spectra. Methods. We carried out analytical calculations based on simple toy models. Simulations of the most extreme situations identified were used to highlight these problems and corrections. Archival VLT/MUSE data of the young PDS70 and HTLup systems were used to vet the detection and characterization capabilities using on-sky observations. New images of the YSES1 planetary system were used to further illustrate the gains. Results. We show that the state-of-the-art method, based on low-pass filtering, can lead to the self-subtraction of the emission lines and modify the neighboring continua. We show that the proposed technique corrects these characterization problems, while maintaining the same detection capabilities. The two protoplanets PDS70 b and c were detected with 5sigma significance. The Halpha line estimate of the HTLup B stellar companion was improved by ~30% for the integrated flux and by ~8% for the 10%-width. As for YSES1 b, we found it uniquely displays a combination of Halpha, Hbeta, CaII H&K triplet, and HeI lines in emission. Conclusions. The proposed method better preserves the spectral information, notably the emission line fluxes and profiles, while achieving similar detection performance. Based on a linear and parametric approach, it can be extended and/or combined with additional faint signal search algorithms.

Figures (23)

• Figure 1: Toy model with rectangular line and flat neighboring continuum.
• Figure 2: Signal processing steps leading to post-subtraction planetary spaxel estimates. (a) Data spaxel ratio with regard to the reference spectrum. (b) Low-pass filtering with Savitzky-Golay filter. (c) Stellar spaxel estimation with overfitting. (d) Planetary spaxel estimation with self-subtraction.
• Figure 3: Subtraction results at the position of a planet with the SGF method. Left: Overlapping simulation (ground truth) and estimation of the stellar spaxel. There is no emission line in this simulation. Middle: The green spaxel is the estimation of the planetary spaxel, while the cyan one is the associated ground truth. The self-subtraction problem is clearly visible in this situation. Right: The reason for the self-subtraction is evidenced. This is the deformation estimation, in light pink, overfitting the planetary line component of the data to stellar spectrum estimation ratio, in purple. Yet, this overfitting is relatively greater than the planetary and stellar continuum-to-line ratios are different. The real deformation is in dark pink.
• Figure 4: Same as Fig. \ref{['fig:sgf_with_out_line']}, but with a small stellar line (total flux of $5\times10^{-13}$$\text{erg}.\text{s}^{-1}\text{cm}^{-2}$Å$^{-1}$ in the 6560.3-6565.3Å band).
• Figure 5: Same as Fig. \ref{['fig:sgf_with_small_line']}, but with a brighter stellar line (total flux of $5\times10^{-12}$$\text{erg}.\text{s}^{-1}\text{cm}^{-2}$Å$^{-1}$ in the 6560.3-6565.3Å band).