Transmission spectroscopy of WASP-52 b with JWST NIRISS: water and helium atmospheric absorption, alongside prominent star-spot crossings

ABSTRACT

In the era of exoplanet studies with JWST, the transiting, hot gas giant WASP-52 b is an excellent target for atmospheric characterization through transit spectroscopy. WASP-52 b orbits an active K-type dwarf recognized for its surface heterogeneities, such as star-spots and faculae, which pose challenges to atmospheric characterization via transmission spectroscopy. Previous transit observations have detected active regions on WASP-52 through crossing events in transit light curves and via the spectral imprint of unocculted magnetic regions on transmission spectra. Here, we present the first JWST observations of WASP-52 b. Our Near Infrared Imager and Slitless Spectrometer/Single Object Slitless Spectroscopy (NIRISS/SOSS) transit observation, obtained through the GTO 1201 programme, detects two clear spot-crossing events that distort the 0.6–2.8 |$\mu$|m transit light curves of WASP-52 b. We find that these two occulted spots cover together about 2.4 per cent of the stellar surface and have temperatures about 400–500 K cooler than the stellar photosphere. Our NIRISS/SOSS transmission spectrum is best-fit by an atmosphere with H|$_2$|O (10.8|$\sigma$|⁠), He (7.3|$\sigma$|⁠, with evidence of an escaping tail at |$\sim$|2.9|$\sigma$|⁠), hints of K (2.5|$\sigma$|⁠), and unocculted star-spots and faculae (3.6|$\sigma$|⁠). The retrieved H|$_2$|O abundance (⁠|$\log$| H|$_2$|O |$\approx -4 \pm 1$|⁠) is consistent with a subsolar or solar atmospheric metallicity for two independent data reductions. Our results underscore the importance of simultaneously modelling planetary atmospheres and unocculted stellar heterogeneities when interpreting transmission spectra of planets orbiting active stars and demonstrate the necessity of considering stellar contamination models that account for both cold and hot active regions.

1 INTRODUCTION

Transmission spectroscopy has proven to be one of the most powerful techniques for probing the composition of exoplanet atmospheres and making inferences about the formation and migration pathways of exoplanets (Seager & Sasselov 2000; Brown 2001). JWST is now the leading observatory for characterizing the atmospheres of exoplanets through transmission spectroscopy (e.g. Fisher et al. 2024). The observations made with this new telescope have been fruitful so far, demonstrating exquisite precision, revealing several spectral features, and leading to the first detections of CO₂ and SO₂ in the atmosphere of a hot Jupiter (WASP-39 b; Alderson et al. 2023; JWST Transiting Exoplanet Community Early Release Science Team 2023; Rustamkulov et al. 2023). Beyond providing enhanced precision, JWST also expands on the spectral range of its predecessors and now covers the absorption features of the most abundant molecules expected in hot Jupiter atmospheres, such as H₂O, CH₄, CO, and CO₂, as well as alkali metals like Na and K (e.g. Burrows & Sharp 1999). With sensitivity to such a wide range of molecules, JWST offers an opportunity to study many out-of-equilibrium effects that occur in exoplanet atmospheres, such as vertical mixing and photochemistry (e.g. Tsai et al. 2023; Welbanks et al. 2024). Additionally, JWST can provide new insights into the presence and characteristics of hazes and clouds (e.g. Bell et al. 2024; Inglis et al. 2024). A wide range of exotic condensates has been proposed to exist, ranging from pure transition metals (like Fe) to silicates, sulphides, and salt/alkali condensates, which may give rise to significant cloud coverage (e.g. Wakeford & Sing 2015). However, clouds can also complicate transmission spectroscopy analyses by covering considerable portions of the atmosphere, making it challenging to identify the chemical species present therein (e.g. Deming et al. 2013). Stellar contamination has proven to be another major challenge in studying the atmospheres of exoplanets, limiting inferences for rocky planets (e.g. Lim et al. 2023; Moran et al. 2023; Radica et al. 2025) and potentially biasing those of inflated gas giants if not well accounted for (e.g. Barstow et al. 2015; Fournier-Tondreau et al. 2024).

The inflated (⁠|$R$| = 1.27 |$R_J$|⁠), Saturn-mass exoplanet (⁠|$M$| = 0.46 |$M_J$|⁠) WASP-52 b (Hébrard et al. 2013) is an ideal target for atmospheric studies given its deep transit (⁠|$\delta$| = 2.71 per cent), and large scale height (⁠|$H$|  |$\approx$| 700 km), which result from its high temperature (⁠|$T_{\rm eq}$| = 1315 K; orbital period of about 1.75 d) and low surface gravity (⁠|$\log g_p$| = 2.81). Prior analyses have often found muted spectral features in transmission spectra of WASP-52 b, which have been attributed to a cloudy atmosphere (Kirk et al. 2016; Chen et al. 2017; Alam et al. 2018). Nonetheless, water vapour absorption has been observed in the near-infrared (Bruno et al. 2018, 2020; Tsiaras et al. 2018), whereas sodium, potassium, hydrogen, and helium have been detected at high-resolution (Chen et al. 2020; Kirk et al. 2022; Canocchi et al. 2024).

The effect of its young and active K2V host star has been carefully considered in previous transit studies (e.g. Kirk et al. 2016; Bruno et al. 2020). Stellar activity has long been acknowledged as a significant obstacle to characterizing exoplanet atmospheres through transmission spectroscopy (e.g. Pont et al. 2008; Czesla et al. 2009). The presence of stellar heterogeneities, such as star-spots and faculae, can cause bumps or dips in light curves when situated along the transit path (e.g. Pont et al. 2007). Outside the transit chord, stellar heterogeneities cause a mismatch between the light source sampled by the planet’s transit and the assumed full stellar disc spectrum, resulting in the so-called transit light source effect (TLSE; Rackham, Apai & Giampapa 2018). Therefore, occulted active regions can hinder the correct measurement of transit parameters and depths by substantially impacting transit light curves (e.g. Barros et al. 2013; Oshagh et al. 2014), whereas unocculted active areas can introduce spurious spectral features by the TLSE (e.g. McCullough et al. 2014; Rackham et al. 2018).

Many efforts have been made to mitigate the impact of stellar activity and achieve unbiased atmospheric inferences (see Rackham et al. 2023 for a recent review with current recommendations). For occulted active regions, simply masking them when fitting light curves does not completely negate their effect; accounting for them with a spot-transit model is the recommended strategy to mitigate their impact (Rackham et al. 2023); otherwise, these magnetic regions can potentially imprint strong slopes at short visible wavelength (e.g. Oshagh et al. 2014). Previous studies of WASP-52 b have revealed transit light-curve anomalies associated with occultations of star-spots (e.g. Mancini et al. 2017; Bruno et al. 2018; May et al. 2018) and a facula (Kirk et al. 2016). When modelling active regions, one can parametrize their position and size on the stellar surface using a ‘white’ light curve and then infer their temperature from the spectrophotometric light curves. However, degeneracies exist between all these parameters (Rackham et al. 2023; Fournier-Tondreau et al. 2024), not only between size and temperature (e.g. Pont et al. 2008). One approach worth considering, which we explore here, is the simultaneous and joint fitting of all spectrophotometric light curves using both wavelength-dependent and wavelength-independent parameters. This method circumvents the need to fix the positions and sizes of active regions based on a broad-band light-curve fit, thereby better exploiting the wavelength-dependence effect of stellar heterogeneities.

For unocculted active regions, the current state of the art is to jointly fit the properties of the stellar heterogeneities and the planetary atmosphere when performing retrievals on transmission spectra (e.g. Pinhas et al. 2018; Rathcke et al. 2021). For WASP-52 b, Bruno et al. (2020) presented a joint retrieval analysis of its optical to infrared transmission spectrum combining HST  ¹/WFC3² (Bruno et al. 2018), HST/STIS³, and Spitzer/IRAC⁴ (Alam et al. 2018) observations.They found unocculted star-spots covering 5 per cent of the stellar surface with a temperature <3000 K, and reported a water abundance of |$\log$| H₂O = |$-3.30^{+0.94}_{-1.12}$|⁠, a |$\sim$|0.1–10|$\times$| solar metallicity and a subsolar carbon-to-oxygen ratio (C/O). Earlier studies (e.g. Alam et al. 2018) instead directly corrected the transmission spectrum based on stellar activity monitoring or occulted heterogeneity properties; however, this approach was proposed to underestimate the impact of unocculted active regions (Rackham et al. 2023). Hot giants can ultimately serve as test cases for validating and improving stellar contamination mitigation since they exhibit absorption features with similar amplitudes to those imprinted by stellar heterogeneities, as opposed to terrestrial planets, where TLSE can dominate over atmospheric features.

Here, we present the 0.6–2.8 |$\mu$|m transmission spectrum of WASP-52 b, obtained with the Near Infrared Imager and Slitless Spectrometer (NIRISS) instrument aboard JWST. Spot-crossing events deform the transit light curves. We thus model these active regions in the light-curve fitting process and then jointly fit the properties of the planetary atmosphere and the unocculted stellar heterogeneities by performing retrievals on the resulting transmission spectrum. We briefly outline the observations and data reduction in Section 2. We then describe the light-curve fitting and the modelling of the spot-crossing events in Section 3. We detail the retrieval analysis in Section 4 and follow this with the helium absorption analysis in Section 5. We summarize and discuss the results in Section 6 and conclude in Section 7.

2 OBSERVATIONS AND DATA REDUCTION

A transit observation of WASP-52 b was obtained using the Single Object Slitless Spectroscopy (SOSS) mode of the NIRISS instrument (Albert et al. 2023; Doyon et al. 2023) as part of the JWST Guaranteed Time Observations (GTO) programme cycle 1 (PID: 1201; PI: David Lafrenière). The time series observation (TSO) started on 2022 November 27, at 07:08:33.169 utc and spanned 4.44 h, which covered the 1.8 h transit as well as 1.9 h of baseline before and 0.74 h after the transit. It employed the standard GR700XD/CLEAR combination, along with the SUBSTRIP256 detector, which captures diffraction orders 1 and 2 of the SOSS mode. A total of 265 integrations were obtained, each consisting of 10 groups and lasting 60.434 s.

We use stages 1 and 2 of the exoTEDRF⁵ pipeline (e.g. Radica et al. 2023; Radica 2024) to reduce the TSO, starting from the raw, uncalibrated files downloaded from the Mikulski Archive for Space Telescopes (MAST). In stage 1, exoTEDRF employs the same steps as the official jwst pipeline, except for the 1/|$f$| noise correction, to perform the detector-level calibrations. However, we do not apply the RefPixStep of the jwst pipeline, as the 1/|$f$| noise correction from exoTEDRF serves the same purpose, including the correction of the even–odd row variations (e.g. Feinstein et al. 2023).

As explained in Radica et al. (2023), the zodiacal background must be subtracted before correcting for the 1/|$f$| noise. Since a constant scaling of the Space Telescope Science Institute (STScI)⁶ SOSS SUBSTRIP256 background model does not completely remove the background (e.g. Lim et al. 2023; Fournier-Tondreau et al. 2024), we separately scale both sides of the pick-off mirror jump of the STScI model. This is done using regions on the top of the detector above the third-order trace, redwards (⁠|$x \in [228,250],~ y \in [380,574]$|⁠) and bluewards (⁠|$x \in [227,250],~ y \in [793,897]$|⁠) of the background jump. The scaling factors between the median frame for each group and the background model are calculated by considering the 16th and 12th percentiles of the distribution of the ratios for the left and right side of the background step, respectively, to account for the small amount of flux present in the two regions.

To remove the 1/|$f$| noise, which is treated at the group level, we carefully mask every order 0 contaminant as well as the two dispersed contaminants (one in the upper left corner and one below the first-order’s trace in the centre) on the detector to make sure they do not bias the reduction. Once the 1/|$f$| correction is done (refer to Radica et al. 2023 for a detailed description of this step), the previously subtracted background is readded to each group in each integration. The final removal of the zodiacal background is performed in stage 2, following the same procedure described above, along with further calibrations such as flat fielding and warm-pixel interpolation.

The 1D spectral extraction is carried out using the ATOCA algorithm (Darveau-Bernier et al. 2022), which accounts for contamination between the first- and second-order on the detector. The APPLESOSS code (Radica et al. 2022) is used to create the specprofile reference file needed for ATOCA. During the extraction, all pixels flagged as DO_NOT_USE and all order 0 contaminants are modelled by the ATOCA algorithm. The spectra are extracted on the decontaminated traces using a box width of 30 pixels. We then use the PASTASOSS package⁷ (Baines et al. 2023) to obtain the wavelength solution for WASP-52 b’s observation. Lastly, any data point deviating by more than 5|$\sigma$| in time was clipped.

A summary of the major reduction steps can be visualized in Fig. A1. Furthermore, we reduced the observations with another independent pipeline, NAMELESS (e.g. Coulombe et al. 2023; Radica et al. 2023, Coulombe et al. 2025), to verify the consistency of our results (see Appendix  B).

3 LIGHT-CURVE FITTING AND OCCULTED STAR-SPOT ANALYSIS

3.1 White light-curve fitting

We construct a broad-band light curve, following Fournier-Tondreau et al. (2024), by summing the flux from wavelengths bluewards of 1.5 |$\mu$|m in order 1 (0.85–1.5 |$\mu$|m) and from 0.65–0.85 |$\mu$|m in order 2 to keep the wavelength range where the spot crossings have a more substantial effect. We mask the first 10 integrations and the 65th, which had an anomalous background signal. The resulting broad-band light curve is shown in Fig. 1 and displays two clear spot-crossing events, seen as bumps near the beginning and end of the transit. We fit a spot-transit model with two spot-crossing events using spotrod (Béky, Kipping & Holman 2014) with the Juliet package (Espinoza, Kossakowski & Brahm 2019). We fix the orbital period to 1.7497798 d and the eccentricity to 0 (Hébrard et al. 2013; all the parameter values used in this paper are listed in Table 1). We fit for the mid-transit time |$t_0$|⁠, the impact parameter |$b$|⁠, the scaled semimajor axis |$a/R_*$|⁠, the scaled planet radius |$R_\mathrm{p}/R_*$|⁠, the spots’ |$x$|- and |$y$|-position, the spots’ radius |$R_\mathrm{spot}$|⁠, the spot-to-stellar flux contrast |$F_\mathrm{spot}/F_*$|⁠, a term to adjust the zero point of the transit baseline |$\rm \theta _0$|⁠, and the two quadratic limb darkening (LD) parameters (⁠|$q_1$|⁠, |$q_2$|⁠) following the parametrization of Kipping (2013). We also fit for a scalar jitter term, |$\sigma$|⁠, which is added in quadrature to the flux error. We test detrending against linear models with time, trace |$x$|-position, trace |$y$|-position, and a linear and quadratic model with time. We find that the broad-band light curve is best-fit by a transit model with a time-dependent slope |$\rm \theta _1$|⁠. We fit 17 parameters and sample the parameter space with 2000 live points using dynesty (Speagle 2020). The priors and the best-fitting transit and spot parameters for the broad-band light-curve fit are shown in Table A1. The reduced chi-squared statistic for the fit with the highest likelihood is |$\chi ^2_\nu$| = 1.18. This best-fitting spot-transit model is overplotted in the top panel of Fig. 1, and a physical representation of the spot crossings is shown in the bottom panel. We also tested a model with a facula (bright region) near the middle of the transit instead of two dark spots by allowing the contrast to be greater than one, but we ruled it out because the model could not reproduce the sharpness of the two bumps or their asymmetry. Furthermore, we assess the H-alpha light curve and do not detect any features that could indicate flares.

3.2 Spectrophotometric light-curve fitting

We proceed to fit the spectrophotometric light curves at a resolving power of |$R$| = 100. At this point, |$t_0$|⁠, |$b$|⁠, |$a/R_*$|⁠, the position and radius of the spots are fixed to the best-fitting values with the highest likelihood from the broad-band fit following Fournier-Tondreau et al. (2024). The remaining transit parameters to be fitted include the scaled planet radius, the contrast for each spot, the two quadratic LD parameters for each spectrophotometric light curve, alongside |$\sigma$|⁠, |$\rm \theta _0$|⁠, and |$\rm \theta _1$|⁠. We put Gaussian priors on the LD parameters based on calculations from the ExoTiC-LD package (Wakeford & Grant 2022) using the 3D stagger grid (Magic et al. 2015). The widths of the Gaussian priors are set to 0.2 following Patel & Espinoza (2022). We use 500 live points for each spectral bin. The spectrophotometric light curves for 14 bins at a resolving power of |$R$| = 100, along with their corresponding best-fitting spot-transit model, are displayed in Fig. 2. The resulting transmission spectrum is shown in Fig. 3.

3.3 Inferred occulted star-spot properties on WASP-52

We constrain the temperature of each occulted star-spot by fitting PHOENIX synthetic stellar atmosphere spectra (Husser et al. 2013) to the retrieved contrast spectrum of each spot. We fit the spot temperature and set the surface gravity of the spot model as a free parameter following Fournier-Tondreau et al. (2024). We model each spot contrast spectrum by taking the flux ratio of a spot spectrum to the star spectrum. For the spot model, we use PHOENIX stellar models with temperatures from 4000 to 5000 K, logarithmic surface gravities (⁠|$\log \, g$|⁠) from 1.5 to 5.5 dex, and a fixed metallicity of 0.03 from Hébrard et al. (2013), and we interpolate these spectra linearly in temperature and |$\log \, g$|⁠. We compute a stellar spectrum for the star model with a temperature, |$\log g$|⁠, and metallicity fixed to Hébrard et al. (2013) values. We employ dynesty (Speagle 2020) with 500 live points to chart the parameter space. Fig. 4 shows for each occulted spot the best-fitting contrast model overplotted on each retrieved spot contrast spectrum.

The broad-band light-curve fit led to precise measurements of the position of the spot occulted after mid-transit but did not provide a well-constrained position for the first occulted spot. Similarly to Fournier-Tondreau et al. (2024), the best-fitting value of the |$y$|-position varies substantially between different broad-band light-curve fits (without changing anything), and is highly correlated with the spot size and contrast (see the corner plot in Fig. A2). Nonetheless, we find two cold spots of sizes |$R\rm _{spot,1}$| = 0.13|$^{+0.05}_{-0.04}$| R|$_{*}$| and |$R\rm _{spot,2}$| = 0.085|$^{+0.03}_{-0.019}$| R|$_{*}$|⁠, with corresponding temperatures of |$\rm \Delta T_1$| = 420 |$\pm$| 20 K and |$\rm \Delta T_2$| = 480 |$\pm$| 20 K cooler than the photosphere. The coverage fraction of these two occulted spots combined is thus about 2.4 per cent of the visible stellar hemisphere. The surface gravities of the spot models are lower than the star's by |$\Delta \log g_1$| = 2.4|$^{+0.4}_{-0.3}$| dex and by |$\Delta \log g_2$| = 1.4|$^{+0.6}_{-0.5}$| dex. These lower |$\log g$| values for the spot models are similar to those inferred for HAT-P-18 (Fournier-Tondreau et al. 2024) but larger than what is physically expected. The increased magnetic pressure in active features decreases the gas pressure, which can be represented by a stellar model with a surface gravity lower by 0.5–1 dex (Solanki 2003; Bruno et al. 2022). Therefore, we test a model that fits only for the spot temperature, fixing the surface gravity to the stellar value. For the second occulted spot, the well-constrained one, we find only weak evidence (log Bayes factor |$\ln \mathcal {B}_{01}$| = 1.3, |$\chi^2 = 153$|⁠) to treat the surface gravity of the spot model as a free parameter, whereas for the first spot, there is strong evidence (⁠|$\ln \mathcal {B}_{01}$| = 6.3, |$\chi^2 = 162$|⁠). Still, that does not impact the retrieved spot temperatures significantly.

Furthermore, we explore another light-curve fitting method that does not rely on a broad-band light-curve fit in an attempt to lift some degeneracies. In this approach, all spectral light-curves are simultaneously and jointly fit by treating some parameters common to all spectral bins (achromatic) and some parameters varying for each spectral bin (chromatic). With this new approach, we retrieve the same solution for the second occulted star-spot and still find degenerate solutions for the |$y$|-position of the first one, where the |$y$|-position is orthogonal to the planet's transit path, leading to different spot sizes and temperatures.

4 RETRIEVAL ANALYSIS

We present inferences from a Bayesian retrieval analysis applied to WASP-52 b’s transmission spectrum. Given the strong signature of occulted star-spots in the spectrophotometric light curves, we additionally consider the influence of unocculted stellar active regions jointly with atmospheric models. In what follows, we first outline our retrieval configuration before detailing our interpretation of WASP-52 b’s transmission spectrum in terms of its planetary atmosphere and the stellar contamination.

4.1 Retrieval configuration

We perform retrievals on WASP-52 b’s transmission spectrum using the open-source retrieval code poseidon (MacDonald & Madhusudhan 2017a; MacDonald 2023), which uses the multinest algorithm to explore the multidimensional parameter space (Feroz, Hobson & Bridges 2009). All retrievals are conducted with 1000 multinest live points to smoothly explore the posterior distributions of the parameter space.

Our atmospheric model assumes an isothermal pressure–temperature (P–T) profile dominated by |${\rm H_2}$| and |${\rm He}$| (with an assumed ratio of |${\rm He}$|/|${\rm H_2}$| = 0.17). The atmospheric model spans |$10^{-7}$|–10|$^2$| bar, with 100 layers spaced uniformly in log pressure, and uses a reference pressure of 10 bar as the boundary condition for hydrostatic equilibrium (i.e. the pressure at which the retrieved reference radius is located). We consider several trace chemical species that are expected to be present at the equilibrium temperature of WASP-52 b (Madhusudhan et al. 2016; Woitke et al. 2018; Mukherjee et al. 2024) and exhibit strong absorption features in the NIRISS/SOSS wavelength range: |${\rm H_2O}$|⁠, |${\rm Na}$|⁠, |${\rm K}$|⁠, |${\rm CO}$|⁠, |${\rm CO_2}$|⁠, |${\rm CH_4}$|⁠, |${\rm NH_3}$|⁠, and |${\rm HCN}$|⁠. Our opacities use state-of-the-art line lists and pressure broadening parameters included in the recent poseidon v1.2 (Mullens, Lewis & MacDonald 2024) release, from the following line list sources: |${\rm H_2O}$| (Polyansky et al. 2018), |${\rm Na}$| and |${\rm K}$| (Ryabchikova et al. 2015), |${\rm CO}$| (Li et al. 2015), |${\rm CO_2}$| (Yurchenko et al. 2020), |${\rm CH_4}$| (Yurchenko et al. 2024), |${\rm NH_3}$| (Coles, Yurchenko & Tennyson 2019), and |${\rm HCN}$| (Barber et al. 2014). The cross-sections for each of these chemical species are calculated by the open-source python package Cthulhu (Agrawal & MacDonald 2024). We also include collision-induced absorption from |${\rm H_2}$|–|${\rm H_2}$| and |${\rm H_2}$|–|${\rm He}$| (Karman et al. 2019) and |${\rm H_2}$| Rayleigh scattering (Hohm 1994). Finally, we follow a parametric treatment of aerosols via the four-parameter inhomogeneous cloud and haze prescription from MacDonald & Madhusudhan (2017a).

Given the evidence of unocculted stellar features from the HST transmission spectra of WASP-52 b (Bruno et al. 2020), we also include stellar contamination parameters in our retrievals. We use three sets of model configurations, as described in Fournier-Tondreau et al. (2024): (i) an atmosphere-only model, (ii) a one-heterogeneity + atmosphere model, and (iii) a star-spots + faculae + atmosphere model. We apply all three models to the exoTEDRF transmission spectrum. We also explore the sensitivity of different data reductions by running both atmosphere-only and star-spots + faculae + atmosphere retrievals on the NAMELESS spectrum. The one heterogeneity model is defined by the stellar photosphere temperature, the heterogeneity temperature (less than the photosphere for spots, greater than the photosphere for faculae), and the heterogeneity covering fraction. The star-spots + faculae model includes two heterogeneities, one assumed to be cooler than the photosphere (spot), and one assumed to be warmer than the photosphere (faculae), each with its own covering fraction. Thus, the one-heterogeneity model adds three free parameters, while the star-spots + faculae model adds five free parameters. We additionally tested retrieving different surface gravities for the heterogeneities compared to the photosphere (as in Fournier-Tondreau et al. 2024) but found this unnecessary. We calculate the contribution from stellar contamination by interpolating PHOENIX models (Husser et al. 2013) using the pymsg package (Townsend & Lopez 2023).

We compute model spectra at a spectral resolution of |$R = \lambda /d\lambda = 20,000$| from 0.58 to 2.84 |$\mu$|m using the configuration described above. The model spectra are calculated via opacity sampling on to this intermediate resolution wavelength grid from the high-resolution opacities (⁠|$\Delta \nu$| = 0.01 cm|$^{-1}$|⁠, equivalent to |$R = \lambda /\Delta \lambda = 10^6$| at 1 |$\mu$|m) described above. We additionally include a relative offset parameter, |$\delta _{\rm rel}$|⁠, to account for the NIRISS/SOSS order 1 and 2 spectra. Taking into account the atmospheric properties, stellar contamination properties, and order 2 versus 1 offset, our retrieval models have the following number of free parameters: 13 for the atmosphere-only model, 16 for the one-heterogeneity + atmosphere model, and 18 for the star-spots + faculae + atmosphere model. We summarize the priors for each model in Table 2.

4.2 Retrieval results

WASP-52 b’s JWST NIRISS transmission spectrum can be explained by H₂O and K absorption alongside unocculted stellar active regions and atmospheric aerosols. We detect H₂O at 10.8|$\sigma$| confidence and find hints of K at 2.5|$\sigma$| confidence, but find no evidence of any other chemical species. Fig. 5 demonstrates that the NIRISS/SOSS data exhibit a spectral slope with increasing transit depth towards short wavelengths, which can be explained by either a scattering haze or unocculted star-spots (McCullough et al. 2014). When spots (but not faculae) are included in the retrieval model, we find that they partially substitute for the haze as the preferred explanation for the spectral slope in the NIRISS/SOSS data (as discussed further in Section 4.2.2) and increase the offset between the NIRISS/SOSS orders. However, the model that includes spots and faculae (preferred at 3.6|$\sigma$| over the atmosphere-only model and 4.4|$\sigma$| over the spot-only model) still requires a contribution from atmospheric hazes. Regardless of the inclusion of stellar contamination, we find that an atmospheric haze with an inhomogeneous terminator fraction is needed to explain the observations. We explore these atmospheric and stellar inferences quantitatively in the following sections.

4.2.1 Unocculted active regions on WASP-52

We first determine whether unocculted active regions are required to explain WASP-52 b’s NIRISS transmission spectrum. Table 2 (lower rows) summarizes various fit quality and model comparison metrics (reduced chi-squared |${\chi _\nu }^2$|⁠, log Bayesian evidence |$\ln \mathcal {Z}_{\mathrm{Bayesian}}$|⁠, Bayes factors |$\mathcal {B}_{01}$|⁠, and detection significances) from our retrievals using different stellar contamination models and data reductions. Across both transmission spectra (exoTEDRF and NAMELESS), we find that the model including star-spots and faculae has the highest Bayesian evidence with equivalent detection significances (of 3.6|$\sigma$| for exoTEDRF and 3.5|$\sigma$|for NAMELESS compared to the atmosphere-only model). However, we note that the atmosphere-only model (without stellar contamination) still provides a reasonable fit to the NIRISS/SOSS observations (⁠|${\chi _\nu }^2 \approx 1$|⁠) and thus cannot be formally ruled out. Interestingly, there is no improvement in the Bayesian evidence when switching from the atmosphere-only model to the atmosphere + star-spots (without faculae) model. This indicates that adding a spectral slope from star-spots does not provide additional explanatory power compared to an atmospheric haze, so the fitness metrics penalize the additional three parameters introduced by the one-heterogeneity model. However, incorporating faculae into the model provides additional explanatory power beyond what can be achieved by changing atmospheric aerosol properties. The improvement in fit metrics when adding faculae arises primarily from the data points with lower transit depths near 0.6 |$\rm{\mu m}$| (see Fig. 5). Therefore, in what follows, we refer to the atmosphere + star-spots + faculae model as the ‘preferred’ model.

The preferred stellar contamination model indicates the presence of both cold and hot active regions on WASP-52. Our results for exoTEDRF indicate star-spots and faculae covering |$34^{+10}_{-12}$| per cent and |$18^{+15}_{-8}$| per cent of the visible stellar hemisphere, with corresponding temperatures |$\approx$|450 K cooler and |$\approx$|650 K hotter than the stellar photosphere, respectively. We find similar results for NAMELESS, albeit with a lower star-spot covering fraction of |$23^{+11}_{-7}$| per cent (see Table 2). These results support the presence of unocculted active regions on WASP-52, alongside the occulted heterogeneities observed in the transit light curves (see Fig. 1), consistent with previous HST observations (Bruno et al. 2018, 2020).

4.2.2 The atmosphere of WASP-52 b

We next report our atmospheric retrieval inferences and their sensitivity to stellar contamination model choices and data reductions (exoTEDRF versus NAMELESS pipeline). Fig. 5 presents our retrieval results for the exoTEDRF for the three stellar contamination models, while Fig. 6 shows the sensitivity of our retrieval results to the two transmission spectra for the statistically preferred retrieval model (star-spots + faculae). We also provide the full retrieval results across all model and data combinations in Table 2.

Our preferred atmospheric retrieval model favours subsolar-to-solar H₂O and K abundances. We find broadly consistent H₂O abundances for both the atmosphere-only (exoTEDRF, atmosphere-only: |$\log \mathrm{H_2 O} = -3.75^{+0.88}_{-0.57}$|⁠) and atmosphere + star-spots + faculae models (exoTEDRF, star-spots + faculae: |$\log \mathrm{H_2 O} = -4.18^{+0.60}_{-0.44}$|⁠), which are consistent with either a solar (⁠|$\log \mathrm{H_2 O}_{\rm {solar}} \approx -3.3$|⁠; Asplund, Amarsi & Grevesse 2021) or somewhat subsolar atmospheric metallicity. For the preferred retrieval model, we also find excellent agreement between exoTEDRF (star-spots + faculae: |$\log \mathrm{H_2 O} = -4.18^{+0.60}_{-0.44}$|⁠) and NAMELESS (star-spots + faculae: |$\log \mathrm{H_2 O} = -3.90^{+1.41}_{-0.56}$|⁠). Similarly, we find subsolar K abundances (e.g. |$\log \mathrm{K} = -9.40^{+0.87}_{-0.90}$| for exoTEDRF under the star-spots + faculae model, versus |$\log \mathrm{K}_{\rm {solar}} \approx -6.9$|⁠; Asplund et al. 2021). While we do not detect other gases, our retrievals place strong 2|$\sigma$| upper limits on the abundances of CO₂, CH₄, and NH₃, ruling out abundances exceeding 10 ppm (see Table 2).

Atmospheric hazes also play a role in shaping WASP-52 b’s NIRISS transmission spectrum. Our retrievals consistently favour a strongly scattering haze |$\sim\!\! 10^6$| stronger than H₂ Rayleigh scattering with a scattering power-law exponent of |$\sim\!\! -6$| (see Table 2). The hazes are distributed inhomogeneously around WASP-52 b’s terminator with a covering fraction of |$\approx \!50~{{\ \rm per\ cent}}$| (e.g. |$\phi = 0.49^{+0.08}_{-0.07}$| for exoTEDRF under the star-spots + faculae model). We do not find evidence for an optically thick high-altitude cloud deck for our preferred retrieval model with both star-spots and faculae.

Under the assumption of only unocculted star-spots (without faculae), we find an unphysical atmospheric solution with a significantly higher H₂O abundance than expected for a hot Jupiter like WASP-52 b (⁠|$\log \mathrm{H_2 O} \sim -1$|⁠, or 10 per cent H₂O; see Fig. 5). Besides the high H₂O abundance, this solution requires a lower atmospheric temperature (⁠|$\approx$|900 K versus |$\approx$|1100 K for the star-spots + faculae model), a large |$301^{+94}_{-89}$| ppm offset between the NIRISS orders, and star-spots that are over 1000 K cooler than the stellar photosphere covering |$5^{+3}_{-2}$| per cent of the stellar disc. We regard this extreme solution as unlikely – arising from the complex degeneracy between atmospheric hazes, unocculted star-spots, and the relative offset between the NIRISS orders – with the lower retrieved offsets for the preferred star-spots + faculae model (⁠|$148^{+85}_{-82}$| ppm for exoTEDRF and |$111^{+98}_{-92}$| ppm for NAMELESS) providing additional evidence that both star-spots and faculae must be considered simultaneously to obtain reliable atmospheric inferences.

5 DETECTION OF EXCESS HELIUM ABSORPTION

We search for evidence of upper-atmosphere helium absorption by independently analysing the pixel-resolution light curves and transmission spectrum around 1.083 |$\rm{\mu m}$|⁠. Due to the escaping nature of this atmospheric tracer, which can lead to potential pre- and post-transit absorption, it is essential to analyse the line using a data-driven approach.

We first analyse the light curves at the pixel resolution of NIRISS/SOSS for the 47 pixels around the helium pixel (centred at 1.083 |$\rm{\mu m}$|⁠), covering the 1.062–1.105 |$\rm{\mu m}$| wavelength range. We construct a reference light curve surrounding the helium triplet by averaging all the light curves (a total of 38) from 1.062 to 1.078 |$\rm{\mu m}$| and from 1.088 to 1.105 |$\rm{\mu m}$|⁠. Therefore, this reference light curve is not biased by escaping helium. A linear trend is visible in the out-of-transit baseline of the 47 individual and reference light curves. We thus fit this trend using the reference light curve and subtract it from all the light curves, assuming that no significant variation occurs in the amplitude of the linear trend over this narrow wavelength range. Then, each light curve is normalized to its average out-of-transit flux, measured from exposures at phases < −0.04, which reduces the impact of any extended helium atmosphere beyond transit on the final helium light curve. We then compute relative light curves by dividing each by the detrended reference light curve to extract the excess absorption at each pixel; the result is shown in Fig. 7. We observe clear absorption of |$\sim$|2000 ppm during the transit, and additional post-transit absorption evidenced at |$\sim$|2.9|$\sigma$| (measured as the average flux post transit), indicating potential atmospheric escape beyond the Roche lobe in the form of a cometary-like tail. However, the lack of longer post-transit absorption makes it challenging to confirm the presence of a cometary-like tail and estimate its duration. As a verification, we inspect the surrounding relative light curves of the helium pixel, but they do not show any detectable absorption.

In a second step, we derive the transmission spectrum by measuring the absorption between transit contact points t₂ and t₃ for the 47 relative light curves, which is shown in Fig. 8. We confirm a clear signal of 1916 |$\pm$| 264 ppm (7.3|$\sigma$|⁠) by fitting a Gaussian with a free amplitude and full width at half-maximum (FWHM). Based on the work conducted in Fournier-Tondreau et al. (2024), Radica et al. (2024), and Piaulet-Ghorayeb et al. (2024), we also consider the following model: a Gaussian with a fixed width (0.75 Å) convolved at the native resolution of NIRISS/SOSS (⁠|$R$| = 700), and at the expected location of the He i triplet absorption (10833.33 Å). This model provides an estimate of the resolved helium signature at high-spectral resolution. The best-fitting value for the helium amplitude expected at high-resolution is 4.1 |$\pm$| 0.8 per cent. This signal is easily within reach of current high-resolution spectrographs for more robust detection and interpretation of the helium line shape, further constraining the mass-loss rate and dynamics. As shown in Fu et al. (2022), modelling the unresolved helium triplet with JWST leads to significant degeneracy between the mass-loss rate and the thermospheric temperature. Thus, the optimal approach to comprehensively understanding WASP-52 b's upper atmosphere is to combine JWST analysis with a high-resolution spectrograph to fully resolve the signal both temporally and spectrally – an effort beyond the scope of this paper.

Kirk et al. (2022) were the first to report the detection of helium in WASP-52 b’s atmosphere at high-resolution, whereas Vissapragada et al. (2020) had previously reported an upper limit at low resolution. Kirk et al. (2022) reported an excess absorption of 3.44 |$\pm$| 0.31 per cent with the near-infrared, high-resolution spectrograph (NIRSPEC) instrument on the Keck II telescope and derived a mass-loss rate of |$\sim$| 1.4 |$\times$| 10|$^{11}$| g s|$^{-1}$| using the p-winds code (Dos Santos et al. 2022). More recently, Allart et al. (2023) with the high-resolution, near-infrared spectropolarimeter SPIRou on the Canada–France–Hawaii Telescope (CFHT), found an upper limit at 1.69 per cent that contradicts previously published results. Our detection agrees with Kirk et al. (2022), which contrasts with Allart et al. (2023). Given that WASP-52 is a young active star with intense extreme ultraviolet (XUV) irradiation (Allart et al. 2023), it might be possible that the helium signature of WASP-52 b varies over time due to changes in its stellar environment. To assess the impact of stellar contamination on our measured signature, we follow the methods described in Spake et al. (2018). We assume a helium equivalent width for the stellar line of 0.284 Å (from SPIRou observations; Allart et al. 2023) and the NIRISS spectral bin size of 9.35 Å. We retrieve an upper limit on the stellar contamination of |$\sim$|800 ppm. Therefore, the helium signature cannot be entirely explained by stellar contamination and must have a planetary origin. Additional high-resolution and JWST observations are required to comprehensively characterize the escaping atmosphere of WASP-52 b.

6 SUMMARY AND DISCUSSION

We presented here the first JWST transmission spectrum of WASP-52 b, a hot Saturn-mass exoplanet orbiting an active K dwarf. Our NIRISS/SOSS observations required a careful analysis of stellar active regions, both in terms of crossing events during light-curve fitting and contamination of the 0.6–2.8 |$\rm{\mu m}$| transmission spectrum during atmospheric retrieval. Our main results are as follows:

• We detect two spot-crossing events with NIRISS/SOSS. We thus extract the transmission spectrum using a spot-transit model to infer the properties of the occulted active regions on the host star WASP-52. The two occulted star-spots, one before mid-transit and another after, cover together about 2.4 per cent of the stellar surface with temperatures |$\approx$|400–500 K cooler than the photosphere. The parameters (position, size, temperature) of the second spot are well constrained, but the |$y$|-position of the first spot is degenerate (orthogonal to the path of the planet’s transit) and correlated with its size and temperature. We find that setting the star-spot surface gravity as a free parameter is unnecessary.

• Our retrieval analysis indicates that a model including unocculted spots and faculae, jointly with a planetary atmosphere, provides the best fit to WASP-52 b’s transmission spectrum. Spots and faculae are detected at 3.6|$\sigma$| significance compared to an atmosphere-only model. The unocculted star-spots and faculae cover |$\approx \, 30 \pm 10\,$| per cent and |$\approx \, 20 \pm 10\,$| per cent of the visible stellar hemisphere, respectively, with corresponding temperatures |$\approx$|450 K cooler and |$\approx$|650 K hotter than the stellar photosphere, respectively. The retrieved temperature of the unocculted star-spots is consistent with the inferred temperature of the occulted star-spots.

• We detect multiple strong atmospheric absorption features caused by H₂O vapour (10.8|$\sigma$|⁠). The retrieved H₂O abundances suggest a subsolar or solar atmospheric metallicity (⁠|$\log$| H₂O = −4.18|$^{+0.60}_{-0.44}$| for the exoTEDRF reduction; |$\log$| H₂O = −3.90|$^{+1.41}_{-0.56}$| for the NAMELESS reduction).

• We additionally find evidence of K absorption (2.5|$\sigma$|⁠) and a spectral slope towards shorter wavelengths consistent with scattering from atmospheric hazes.

• Finally, we detect atmospheric He (7.3|$\sigma$|⁠) with an excess absorption near 1.083 |$\rm{\mu m}$| of 1916 |$\pm$| 264 ppm. A tentative signature of additional post-transit absorption is also present in the light curve surrounding the He triplet (⁠|$\sim$|2.9|$\sigma$|⁠), hinting at potential atmospheric escape.

We proceed to discuss the broader context surrounding our findings and the implications of our results.

6.1 The atmosphere of WASP-52 b in context

Our results confirm the presence of H₂O vapour reported in previous works using the HST (Bruno et al. 2018, 2020; Tsiaras et al. 2018), alongside K and He from ground-based observations (Chen et al. 2020; Kirk et al. 2022). Our best-fitting retrieval model (including star-spots and faculae) yields a H₂O abundance of |$\log$| H₂O = −4.18|$^{+0.60}_{-0.44}$| (exoTEDRF) or |$\log$| H₂O = −3.90|$^{+1.41}_{-0.56}$| (NAMELESS), which is consistent within 1|$\sigma$| with the H₂O abundances reported from previous HST observations (⁠|$\log$| H₂O = |$-3.30^{+0.94}_{-1.12}$| (Bruno et al. 2020) and log H₂O = −4.09 |$\pm$| 0.87 (Tsiaras et al. 2018) – though the latter study did not account for stellar activity). Our most precise H₂O abundance (from the exoTEDRF reduction) is subsolar at 2|$\sigma$| (compared to a solar value of |$\log$| H₂O = −3.3; Asplund et al. 2021), which is suggestive of either a subsolar atmospheric metallicity or a supersolar C/O ratio. We note that the wider uncertainties on the H₂O abundance from the NAMELESS reduction permit a solar H₂O abundance. Ultimately, additional observations at longer infrared wavelengths are necessary to detect other chemical tracers, such as CO₂ and CO, to complete the chemical inventory of WASP-52 b’s atmosphere and allow the study of more complex effects (such as disequilibrium chemistry and vertical mixing) and help constrain the planet’s formation history.

Our inference of a subsolar or solar H₂O abundance for WASP-52 b is consistent with the mass–metallicity trend seen in other hot giant exoplanets. In the Solar system, a clear inverse trend is observed between planetary mass and atmospheric metallicity (using CH₄ as a proxy; Atreya et al. 2018), which is interpreted as evidence for formation via core accretion (Pollack et al. 1996). With a mass similar to Saturn and a host star of solar metallicity ([Fe/H] = 0.03 |$\rm {\pm }$| 0.12; Hébrard et al. 2013), WASP-52 b would be expected to have an atmospheric metallicity near the value of 10 |$\times$| solar inferred from Saturn’s CH₄ abundance (Fletcher et al. 2009; Atreya et al. 2018). However, population analyses of hot giant exoplanets have generally favoured a lower trendline for H₂O abundances compared to CH₄ abundances in the Solar system (e.g. Welbanks et al. 2019; Sun et al. 2024). The most recent exoplanet mass–metallicity trend presented by Sun et al. (2024) predicts an atmospheric O abundance for a 0.46 |$M_{\rm J}$| planet almost identical to the stellar O abundance, which is in good agreement with our subsolar to solar H₂O abundance.

Atmospheric hazes are needed to explain WASP-52 b’s JWST NIRISS transmission spectrum. We also find that haze is required regardless of the inclusion of unocculted star-spots, which can mimic the signature of a scattering haze. Observations of muted spectral features in WASP-52 b’s transmission spectra were previously attributed to an optically thick cloud deck (Kirk et al. 2016; Chen et al. 2017; Alam et al. 2018), but the retrieval analysis of the combined HST/STIS, WFC3, and Spitzer/IRAC transmission spectrum performed by Bruno et al. (2020) was compatible with solutions both with and without clouds. Our JWST retrieval analysis finds no evidence for optically thick high-altitude grey clouds, suggesting that the primary aerosols in WASP-52 b’s upper atmosphere are small particles producing scattering hazes. Ultimately, our findings of haze scattering in WASP-52 b’s atmosphere are compatible with theoretical work on aerosols and condensates, along with observational trends that suggest hot Jupiters with equilibrium temperatures similar to WASP-52 b should have hazy atmospheres (e.g. Barstow et al. 2017).

Additional observations will help us better understand the atmospheric chemical compositions and aerosol properties of WASP-52 b. The JWST GO⁸ 3969 programme has recently obtained a NIRSPec⁹/G395H transit spectrum for this planet ranging from 2.8 to 5.2 |$\mu$|m. Combining the NIRSpec and NIRISS observations will enable detailed constraints on the atmospheric inventory of key O, C, and N-bearing molecules, including |$\rm {HCN}$| near 3.1 |$\rm{\mu m}$| (MacDonald & Madhusudhan 2017b), |$\rm {H_2 S}$| near 4.0 |$\rm{\mu m}$| (Fu et al. 2024), |$\rm {SO_2}$| near 4.1 |$\rm{\mu m}$| (Alderson et al. 2023; Rustamkulov et al. 2023; Tsai et al. 2023), |$\rm {CO_2}$| near 4.3 |$\rm{\mu m}$| (JWST Transiting Exoplanet Community Early Release Science Team 2023), and |$\rm {CO}$| near 4.7 |$\rm{\mu m}$| (Rustamkulov et al. 2023).

6.2 Accounting for the stellar activity of WASP-52

6.2.1 Occulted star-spots impacting transit light curves

In this work, we jointly retrieved the parameters of the occulted active regions and the planetary transit. Two independent light-curve fitting methods effectively constrained the properties of the spot occulted after mid-transit. However, similar to the findings of Fournier-Tondreau et al. (2024), our results for the first occulted spot indicate lingering uncertainties regarding the inferred star-spot properties. Deducing the 2D distribution of heterogeneities from transit light curves is a degenerate problem since varying the spot |$y$|-positions (orthogonal to the transit chord) can yield identical light curves. Given the correlation between the position of the spot and its size and temperature, this can lead to different spot properties. Additionally, certain assumptions within the spot-transit model may hinder some solutions from being ruled out. Specifically, spotrod assumes circular, homogeneous active regions (without umbra and penumbra structures for spots, or elongated shapes for faculae), which follow the star LD law (Béky et al. 2014). Moreover, we explored whether using all spectral channels to fit the spot position and radius would lift some degeneracies. Nevertheless, this method did not provide a unique solution for the first spot, which was also loosely constrained by the standard light-curve fitting method. Fortunately, regardless of the occulted spot correction method, our retrieval results for the derived exoTEDRF and NAMELESS transmission spectra are in good agreement despite using different spot parameters (see Table A1).

LD has a significant interplay with active regions, and their presence on stellar surfaces can effectively modify limb darkening profiles (e.g. Csizmadia et al. 2013; May et al. 2018). Even for homogeneous stars, the impact of the choice of LD law and its parametrization on atmospheric results still requires further investigation. A common approach with JWST observations thus far, including the one adopted here, has been to use the quadratic LD law while sampling the two parameters via Kipping (2013) parametrization (e.g. JWST Transiting Exoplanet Community Early Release Science Team 2023). However, recent studies suggest that these choices can introduce bias in the retrieved transit depths (Coulombe, Roy & Benneke 2024; Keers et al. 2024), highlighting that our understanding of LD effects still needs improvement.

Our work highlights the benefits of modelling star-spots in transit light curves. First, the NIRISS/SOSS light curves are significantly deformed by the spot crossings, and simply masking them would have decreased the observing efficiency. Second, the temperatures inferred for the occulted spots agree with those independently derived from our retrieval modelling of unocculted spots, which provides reassurance that both methods reliably infer the properties of WASP-52’s active regions. Furthermore, the retrieved covering fraction of occulted spots is expected to be much lower than that of unocculted spots, as the proportion of the photosphere swept by the planet’s transit chord is relatively small. Observational data suggest that active K-type dwarfs have an overall spot coverage ranging between 20 and 35  per cent (e.g. Nichols-Fleming & Blackman 2020), which is consistent with our results (⁠|$f_{\rm spot} \approx 30 \pm 10$| per cent for exoTEDRF).

6.2.2 Unocculted active regions contaminating transmission spectra

The enhanced information gained from JWST NIRISS compared to its predecessor enables us to mitigate the size-temperature degeneracy of unocculted active regions, a limitation faced in earlier observations with HST (Bruno et al. 2020). This could explain why the inferred unocculted spot properties for our preferred model (⁠|$\Delta T_{\rm spot} \approx$| 450 K with |$f_{\rm spot} \approx 30 \pm 10$| per cent for exoTEDRF) are not in agreement with those reported in Bruno et al. (2020), who found significantly cooler spots with a lower coverage fraction (⁠|$T_{\rm spot}$| < 3000 K with |$f_{\rm spot}$| = 5 per cent). Recently, Savanov & Dmitrienko (2019) demonstrated, using three independent observational sources, that spot temperatures for active stars with |$T_{\rm {eff}} \sim$| 5000 K are expected to be about 750 K cooler than the stellar photosphere, which agrees with our results, but contrasts the earlier trend reported by Berdyugina (2005). Additionally, Bruno et al. (2020) only accounted for unocculted spots, while our analysis indicates that the preferred retrieval model favours both unocculted spots and faculae rather than only spots. This adds to the growing body of evidence that employing a single population of unocculted active regions may be an overly simplified prescription of surface heterogeneities (e.g. Zhang et al. 2018; Lim et al. 2023; Fournier-Tondreau et al. 2024).

In this work, we found that the treatment choice for unocculted active regions can strongly impact the retrieved atmospheric properties. Specifically, adopting a model with only a single stellar heterogeneity biased our retrieved H₂O abundances to unphysically high values via a complex degeneracy between unocculted star-spots, atmospheric hazes, and the atmospheric mean molecular weight. Unlike the findings of Fournier-Tondreau et al. (2024) for HAT-P-18 , we did not find evidence that setting the surface gravities of active regions as free parameters provided any advantage. However, we stress that these joint stellar and planetary retrievals rely on grids of 1D stellar models, which in turn limits their accuracy to the precision of these models (e.g. Iyer & Line 2020).

6.2.3 Implications for stellar contamination modelling in exoplanet atmospheric retrievals

Stellar contamination has rapidly become a key consideration in atmospheric retrievals with the advent of JWST. In particular, searches for rocky planet atmospheres have been hindered by the challenges of stellar contamination (Lim et al. 2023; Moran et al. 2023; Radica et al. 2025). WASP-52 b provides an ideal case study to assess methods for disentangling stellar contamination from atmospheric features, as both effects produce signals of similar magnitudes in WASP-52 b’s transmission spectrum. Our results demonstrate that multiple stellar contamination models, including independent populations of cold and hot active regions, should be considered in atmospheric retrievals.

7 CONCLUSION

The hot gas giant WASP-52 b stands out as a key benchmark for atmospheric characterization and stellar contamination mitigation in the JWST era. In this work, we presented its first JWST observation while addressing the challenges posed by the host star’s activity in characterizing the exoplanet’s atmosphere. Our NIRISS/SOSS observation supported previous water, helium, and potassium detections in WASP-52 b’s atmosphere and the indications of a subsolar or solar atmospheric metallicity. We identified both occulted star-spots and unocculted active regions (both spots and faculae) and emphasize the necessity of accounting for stellar activity when performing atmospheric characterization through transmission spectroscopy of exoplanets orbiting active stars. Specifically, our results highlight the importance of exploring different stellar contamination models to ensure accurate atmospheric inferences and avoid potential biases.

Additionally, our study points to promising avenues for future research. First, the upcoming analysis of the JWST NIRSpec observation will expand the chemical inventory of WASP-52 b and refine constraints on its atmospheric composition and aerosol properties. Second, the degeneracies between stellar contamination and atmospheric parameters underscore the need for continued development of robust models and mitigation strategies to improve exoplanet characterization.

ACKNOWLEDGEMENTS

This work is based on observations made with JWST. This project was undertaken with the financial support of the Canadian Space Agency. This research was enabled in part by support provided by Calcul Québec (www.calculquebec.ca) and the Digital Research Alliance of Canada (alliancecan.ca). This work benefited from financial support from the Fonds de Recherche du Québec–Nature et technologies (FRQNT) and the Natural Sciences and Engineering Research Council (NSERC). MFT acknowledges financial support from the Clarendon Fund Scholarship and FRQNT. KM acknowledges financial support from FRQNT. RJM acknowledges support from NASA through the NASA Hubble Fellowship grantHST-HF2-51513.001, awarded by the STScI, which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contractNAS 5–26555. RA acknowledges support from the Swiss National Science Foundation (SNSF) through the Post-Doc Mobility grantP500PT_222212 and from the Institut Trottier for Research on Exoplanets (iREx). CPG acknowledges support from the E. Margaret Burbidge Prize Postdoctoral Fellowship from the Brinson Foundation. SP acknowledges financial support from the Swiss National Science Foundation (within the framework of the National Centre of Competence in Research PlanetS funded under grant 51NF40_205606). LD acknowledges support from NSERC and the Trottier Family Foundation. NBC acknowledges support from an NSERC Discovery Grant, a Tier 2 Canada Research Chair, and an Arthur B. McDonald Fellowship. NBC also thanks the Trottier Space Institute and iREX for their financial support and for providing a dynamic intellectual environment. DJ is supported by the National Research Council (NRC) Canada and an NSERC Discovery Grant.

DATA AVAILABILITY

All data used in this study are publicly available from the Barbara A. MAST.¹⁰

Footnotes

References

APPENDIX A: ADDITIONAL MATERIALS

We provide supplementary materials related to the data reduction process (data products shown in Fig. A1) and the broad-band light-curve fit (priors and results detailed in Table A1, along with the corner plot displayed in Fig. A2).

APPENDIX B: ADDITIONAL REDUCTION

We perform an independent reduction on the TSO using NAMELESS; we follow the steps described in detail in Coulombe et al. (2023, 2025) to correct for instrumental effects. Bad pixels are flagged using the spatial derivative of the detector images and corrected using bicubic interpolation. The background is subtracted by independently scaling both sides of the background model provided by STScI. Cosmic ray events are addressed by computing the running median over time for all individual pixels and clipping any counts more than 4|$\sigma$| away from their median. We then correct for 1/|$f$| noise by scaling each column of the trace independently (considering only pixels within a 30-pix distance from the centre of the trace, while also independently scaling orders 1 and 2) and subtracting the additive constant that minimizes the chi-squared of the fit between the column at a given integration and its median over time. Background contaminants are masked when treating the 1/|$f$| noise to avoid biasing the light curves. Finally, we extract the light curves using a box aperture with a width of 32 pix and use the wavelength solution from MAST.

The light-curve fitting is done as described in Section 3, except that the flux errors reported by the reduction pipeline are not used; we fit for the scalar jitter term. Table A1 shows the best-fitting broad-band light-curve parameter values. This retrieved transmission spectrum and the one from the reference exoTEDRF reduction are displayed in Fig. B1. These two transmission spectra are in good agreement and display consistent features across the entire wavelength range of NIRISS/SOSS. There is an offset of 174.1 ppm between them, mainly due to the slight differences in the values fixed from the broad-band light-curve fit.

Author notes