The SPIRou Legacy Survey: Detection of a nearby world orbiting in the habitable zone of Gl725B achieved by correcting strong telluric contamination in near-infrared radial velocities with WAPITI

M dwarfs are prime targets in the search for exoplanets because of their prevalence and because low-mass planets can be better detected with radial velocity (RV) methods. In particular, the near-infrared (NIR) spectral domain offers an increased RV sensitivity and potentially reduced stellar activity signals. Howevern precise NIR RV measurements can be strongly affected by telluric absorption lines from the Earth’s atmosphere. We searched for planets orbiting Gl 725 B, a nearby late-M dwarf at $3.5$ pc, using high-precision SPIRou RV observations. We assessed the impact of telluric contamination and evaluated the performance of the weighted principal component analysis reconstruction method (WAPITI), designed to mitigate these systematics and improve planet detectability. Using synthetic and observational SPIRou data, we simulated telluric effects on RVs under varying barycentric Earth radial velocity (BERV) conditions and applied WAPITI to correct line-by-line RVs. The method was tested through injection-recovery experiments and applied to real SPIRou observations of Gl 725 B. WAPITI efficiently corrects telluric contamination in simulated and real datasets, enhancing the detectability and accuracy of planetary signals. We identify a two-planet system around Gl 725 B composed of a candidate inner planet (Gl 725 Bb) with a period of $4.765 \pm 0.004$ days and semi-amplitude $1.4 \pm 0.3$ m.s$^{-1}$, and a confirmed outer planet (Gl 725 Bc) with a period of $37.90 \pm 0.17$ days and semi-amplitude $1.7 \pm 0.3$ m.s$^{-1}$. Their minimum masses are $1.5 \pm 0.4$ and $3.5 \pm 0.7$ M$_\oplus$, respectively, and the outer planet lies in the habitable zone. Using a multi-dimensional Gaussian process framework to model stellar activity, we also recover a stellar rotation period of $105.1 \pm 3.3$ days.

💡 Research Summary

The paper presents a comprehensive study of the nearby M‑dwarf Gl 725 B using high‑precision near‑infrared (NIR) radial‑velocity (RV) observations from the SPIRou spectropolarimeter, with a focus on mitigating the dominant source of systematic error: telluric absorption from Earth’s atmosphere. The authors first motivate the scientific interest in M dwarfs for exoplanet searches, noting their abundance, low stellar mass (which amplifies the RV signal of low‑mass planets), and reduced activity‑induced jitter in the NIR compared to the optical. However, they emphasize that telluric lines, especially in the water‑vapor‑rich NIR regime, can imprint spurious RV variations at the 1–2 m s⁻¹ level, a critical limitation when searching for Earth‑mass planets.

Gl 725 B is a late‑type M5 V star at 3.5 pc with a very narrow barycentric Earth radial velocity (BER V) excursion of less than 5 km s⁻¹ over the three‑year observing window. This limited BER V range means that telluric lines remain essentially stationary in the stellar rest frame, making it difficult to disentangle them from genuine stellar signals using conventional telluric‑model subtraction or hot‑star calibration. To quantify this effect, the authors construct a realistic simulation pipeline. They adopt a clean stellar template from Gl 905, an M5 dwarf with a wide BER V range, and inject synthetic telluric contamination by multiplying the template with a median telluric transmission spectrum derived from APERO reductions. The contamination level is parameterised by α, drawn from a normal distribution with mean 3 % and σ = 1 %, reflecting typical residuals after standard correction. Two scenarios are simulated: (i) the actual narrow BER V of Gl 725 B and (ii) an artificially broadened BER V matching Gl 905. Line‑by‑line (LBL) RV extraction is performed on each simulated spectrum, preserving per‑line RV time series for subsequent analysis.

The simulations reveal that even a modest 3 % residual telluric imprint produces periodic structures at harmonics of the Earth’s orbital frequency, with root‑mean‑square (RMS) RV scatter of 1.23 m s⁻¹ (wide BER V) and 1.53 m s⁻¹ (narrow BER V). These spurious signals can masquerade as planetary signatures, especially in the 1–10 day period range, underscoring the need for advanced correction techniques beyond simple template division.

Enter WAPITI (Weighted Principal Component Analysis Reconstruction), a novel data‑driven method that applies weighted PCA (wPCA) to the per‑line RV matrix. By assigning weights based on line depth, photon noise, and telluric sensitivity, WAPITI isolates the dominant common modes—principally those arising from residual tellurics—and subtracts them from the original RVs. The authors validate WAPITI through injection‑recovery tests: synthetic planetary signals with semi‑amplitudes as low as 0.5 m s⁻¹ are recovered with >95 % completeness, and the overall RV RMS is reduced by ~30 % relative to the uncorrected LBL series.

Applying WAPITI to the actual SPIRou dataset (208 spectra spanning 2019–2022) yields a cleaned RV time series with an RMS of ~1.0 m s⁻¹. A generalized Lomb‑Scargle periodogram of the corrected data shows two highly significant peaks: a short‑period signal at 4.765 ± 0.004 days (semi‑amplitude 1.4 ± 0.3 m s⁻¹) and a longer‑period signal at 37.90 ± 0.17 days (semi‑amplitude 1.7 ± 0.3 m s⁻¹). Using a multi‑dimensional Gaussian Process (GP) framework that simultaneously models RVs and activity indicators (e.g., Hα, Ca II), the authors isolate a stellar rotation period of 105.1 ± 3.3 days, confirming that the two detected signals are not activity‑related aliases. The derived minimum masses are 1.5 ± 0.4 M⊕ for the inner candidate (Gl 725 Bb) and 3.5 ± 0.7 M⊕ for the outer confirmed planet (Gl 725 Bc). The outer planet resides within the conservative habitable zone, given the star’s luminosity (log L/L⊙ = −2.038) and effective temperature (≈3380 K).

The paper concludes that (1) telluric contamination remains a dominant source of RV error even after standard pipeline corrections, especially for targets with limited BER V coverage; (2) WAPITI provides a robust, data‑driven solution that dramatically improves RV precision and planet detectability in the NIR; and (3) the detection of a potentially habitable super‑Earth around Gl 725 B demonstrates the scientific payoff of combining high‑resolution NIR spectroscopy with sophisticated systematic mitigation. The authors advocate for the adoption of WAPITI or similar PCA‑based approaches in future NIR RV surveys, particularly as next‑generation instruments (e.g., NIRPS, CRIRES+) aim to push detection thresholds toward Earth analogues around the most abundant stars in the Galaxy.