betaFeedback

WorksPhysics and Astronomy

The JWST Sub-Jupiters Survey: Observed Sensitivities to the Wide-Separation, Cold, Sub-Jupiter Exoplanet Population

Abstract

We present a sensitivity analysis of three recent JWST direct imaging programs (GO-4050, GO-5835, and SURVEY-6005) designed to build a demographic study of sub-Jupiter mass (0.1−10.1 - 1 MJ_{\mathrm{J}}) exoplanets in nearby young associations and moving groups. We find that these programs are sensitive to sub-Saturn to Jupiter mass exoplanets with effective temperature (Teff_\mathrm{eff}) as low as ≲250\lesssim 250 K, already at ∼2\sim 2 arcsec. A comparison with the SHINE survey confirms that the improvements in mass sensitivity provided by JWST high-contrast imaging surveys outpace current ground observation for masses ≲1\lesssim 1 MJ_{\mathrm{J}} and separations between ∼20−500\sim20-500 au, opening up a novel parameter space to explore. Overall, this analysis provides a robust framework for interpreting future exoplanet detections. By correlating photometrically derived properties (e.g. mass, TeffT_{\mathrm{eff}}, and separation) with theoretical model predictions, we show how the access to potentially low detectable masses (sub-Saturn to Jupiter) strongly depends on the expected atmospheric chemistry of the exoplanet population.

Email: [email protected] provided by JWST high-contrast imaging surveys outpace current ground observation for masses ≲1 MJ\lesssim1\,M_{\mathrm{J}} and separations between ∼20−500\sim20-500 au, opening up a novel parameter space to explore. Overall, this analysis provides a robust framework for interpreting future exoplanet detections. By correlating photometrically derived properties (e.g. mass, TeffT_{\mathrm{eff}}, and separation) with theoretical model predictions, we show how the access to potentially low detectable masses (sub-Saturn to Jupiter) strongly depends on the expected atmospheric chemistry of the exoplanet population.

Intro

Despite the continuous effort from ground-based surveys (e.g.; B. P. Bowler [12]; E. L. Nielsen et al. [61]; R. B. Fernandes et al. [29]; A. Vigan et al. [82]; S. Desidera et al. [27]; B. J. Fulton et al. [32]; V. Squicciarini et al. [77]) and the known exoplanet population rapidly growing every year, empirical constraints on the properties and occurrence rates of low-mass (≤1MJup\leq1M_{\mathrm{Jup}}) wide separation (≳10\gtrsim10 au) giant exoplanets do not exist. Transiting surveys probe the exoplanet population from rocky super-Earths to gaseous super-Jupiters, but are mostly limited to within 1 au separations. Observations through radial velocity (RV) and microlensing techniques push the known population of exoplanets out to ≲10\lesssim10 au separations, but cannot perform follow-up atmospheric characterization. Astrometry can also provides evidence for the presence of exoplanets at closer separation (≲10\lesssim10 au), but solutions at wider separation are still mostly limited to Brown Dwarfs or more massive companions (W. Thompson et al. [81]).

While Gaia DR4 will improve these results, to be able to detect and characterize exoplanets beyond ∼10\sim10 au, we still need to turn to direct imaging techniques. Near-to mid-infrared wavelengths serve as the sweet spot for direct imaging observations. Since young planets retain much of the heat of their primordial formation, and their spectral energy distributions (SEDs) reach their peak, making them relatively warm and luminous at these wavelengths. At the same time, the corresponding SEDs of stellar hosts gradually diminish, and the star-planet contrasts are at a minimum (P. Pathak et al. [64]; S. Hinkley et al. [36]). However, the vast majority of exoplanet imaging observations have so far been performed from the ground where the absorption of Earth’s atmosphere limit our exploration of infrared wavelengths, and those wavelengths that are accessible are strongly affected by the thermal background. In the recent years JWST has changed this paradigm, providing access with direct imaging to a dramatically expanded observable parameter space, as new discoveries of both old and young, cold, exoplanets recently demonstrated (∼275\sim275 K, ∼3.5\sim3.5 Gyr, ∼6.3 MJup\sim6.3\,M_{\mathrm{Jup}} E. C. Matthews et al. [54]; ∼300\sim300 K, ∼4.5\sim4.5 Gyr, ∼6 MJup\sim6\,M_{\mathrm{Jup}} D. C. Bardalez Gagliuffi et al. [7]; ∼275\sim275 K, ∼10\sim10 Myr, ∼0.2−1 MSat\sim0.2-1\,M_{\mathrm{Sat}} A.-M. Lagrange et al. [44]; K. A. Crotts et al. [24]), thanks to its outstanding performance in both infrared sensitivity and wavelength coverage.

Microlensing surveys (D. Suzuki et al. [80]) show evidence for an increase of occurrence rates of 0.05−1.6 MJup0.05-1.6\,M_{\mathrm{Jup}} exoplanets out to 10 au. This is also supported by RV surveys (e.g.; R. A. Wittenmyer et al. [84]) that show no drop-off in the occurrence of >0.3 MJup>0.3\,M_{\mathrm{Jup}} exoplanets out to 10 au. In addition, ALMA observations of substructure in protoplanetary disks seems to suggest occurrence rates of planets between 0.1−1 MJup0.1-1\,M_{\mathrm{Jup}} consistent with results from microlensing surveys S. Zhang et al. [88]. High-precision RV measurements of FGKM stars by B. J. Fulton et al. [32] seems to indicate a peak at 2−52-5 au, and a drop off in the occurrence rate of Jupiter-like exoplanets beyond that. However, if we separate the population of exoplanets into their sub-Jupiter and super-Jupiter components (∼1−20 MJ\sim1-20\,M_{\mathrm{J}}), the picture becomes less clear. When looking at the super-Jupiter population in isolation, the significance of a drop-off at 10 au is much less apparent, and the population seems to plateau. Conversely, the sub-Jupiter population is only measured out to 10 au. Thus it is not clear if the frequency of sub-Jupiter exoplanets plateau or rise past 10 au, or if sub-Jupiter exoplanets are more or less common than super-Jupiter beyond 10 au.

To facilitate answering these questions, the JWST Sub-Jupiters Survey Collaboration was established. Three JWST programs have been performed in the last few years (GO-4050 and GO-5835: A. Carter et al. [19] (2023); and SURVEY-6005: B. Biller et al. [9]) to build a demographic study of sub-Jupiter mass (0.1−1 MJ0.1-1\,M_{\mathrm{J}}) exoplanets in nearby young associations and moving groups (loose, low-density groups of coeval stars). These coeval structures represent one of the best possible environments to search for planetary companions, since being part of an association helps to constrain both the age and the distance of their constituent members (D. R. Soderblom et al. [74]; J. Gagné [33]). Additionally, closer distances provides better access to smaller physical separations, and younger ages provide access to lower mass companions. These programs have already yielded their first results, with the very recent detection of RX J0534.0-0221 b (a ∼2.8 MJ\sim2.8\,M_{\mathrm{J}} planet at ∼14\sim14 au; R. Ferrer-Chavez et al. [30]), and the scheduling of 35 new exoplanet candidate follow-ups for Cycle 5 (GO-10764; A. Carter & B. Biller [17]).

In this paper we discuss the results from the sensitivity analysis performed on the three JWST programs (GO-4050, GO-5835 and SURVEY-6005). § 2 motivates target selection and observation strategy for the 3 JWST programs. In § 3}, we describe the data reduction and analysis techniques used to identify point source candidates and measure sensitivity curves. In Section 4.2 we discuss how we derive the calibrated contrast curves, while in {xref}Section 4.2 we translate these contrast curves into magnitude limits. Finally, in {xref}Section 5 we discuss the survey mass/$T_{\mathrm{eff}}$ sensitivity limits, the detection probability maps, and the expected yield from our three JWST programs combined, before presenting our conclusions in {xref}Section 6`.

Program design

Between June 2024 and September 2025, three JWST NIRCam programs have been completed, targeting a total of 182 targets among nearby young moving groups and associations. A complete summary of their respective observation strategies is presented in Table 1. GO-4050 observed a total of 23 targets plus 7 dedicated reference stars using a coronagraphic setup in filters F200W+MASK335R and F444W+MASK335R. GO-5835 had a total of 45 coronagraphic targets observed in filters F200W+MASK335R and F444W+MASK335R and 11 imaging targets observed in filters F210M and F444W. SURVEY-6005 had a total of 96 coronagraphic targets observed in filters F200W+MASK335R and F444W+MASK335R, of which 10 where later rejected due to coronagraphic misalignment following acquisition. We include all of these targets in our analysis, with the exception of the 7 reference targets from GO-4050, and the 10 misaligned targets from SURVEY-6005. This brings the total number of available targets for the analysis to 165.

Table 1. Survey summary showing the respective GO/SURVEY ID, number of targets, dedicated reference stars, misaligned targets, filter employed, the average charged time per visit, and the beginning and end of the observations.

IDTargetsReferencesMisalignedFiltersHoursStartEnd
40502370F200W, F444W1.58Jun 6, 2024Jun 16, 2024
58355600F200W, F210M, F444W1.86Sep 17, 2024Sep 20, 2025
600586010F200W, F444W1.87Aug 20, 2024Jul 5, 2025
TOTAL1657103331.08

Table 1.

Table 2 shows a breakdown of our targets by their type: if they are either flagged as isolated, hosting a binary or a disk, or both.

Table 2. Targets summary showing for each GO/SURVEY ID, the number of isolated targets, binaries, isolated targets with disks, or both observed per program, and their respective totals.

IDSinglesBinariesDisksBinaries/DisksTotal
40501534123
583532166256
600554265186
TOTAL10145154165

Table 2.

Target selection

We targeted a range of different associations and moving groups (henceforth “group”) as part of our three programs. Table 3 shows a breakdown of the targets by spectral type and group, with the age, the median distance, and relative uncertainties, for our targets in that group, as reported by MOCAdb4 (J. Gagné et al. [34]; J. Gagné [33]): the Beta Pictoris moving group (BPMG: 43 targets, or ∼26%\sim26\% of the total sources), the Tucana-Horologium association (THA: 28 targets, ∼17%\sim17\%), TW Hya association (TWA: 24 targets, ∼14.5%\sim14.5\%), Ursa Major association (UMA: 20 targets, ∼12%\sim12\%), and the AB Doradus moving group (ABDMG: 16 targets, ∼10%\sim10\%), plus a few additional groups with very few targets each: Austral complex (AUSC: 4 targets, ∼2.4%\sim2.4\%), Carina association (CAR: 4 targets, ∼2.4%\sim2.4\%), Carina-Near Moving Group (CARN: 4 targets, ∼2.4%\sim2.4\%), Columba association (COL: 10 targets, ∼6%\sim6\%), FORH-Carina-Columba node of the Austral Complex (FCCN: 1 targets, ≲1%\lesssim1\%), Melange-4 moving group (MEL4: 1 targets, ≲1%\lesssim1\%), Octans association (OCTN: 1 targets, ≲1%\lesssim1\%), Scorpius-Centaurus region (SCOCEN: 1 targets, ≲1%\lesssim1\%), Vulpecula system (VUL: 1 targets, ≲1%\lesssim1\%), and field stars (FIELD: 7 targets, ∼4.2%\sim4.2\%). GO-4050 targeted primarily M stars in TWA association, while GO-5835 targeted primarily M stars in BPMG moving group. In contrast, SURVEY-6005 was a survey-mode program, designed to fill gaps in the JWST observing schedule, and hence observed FGK stars across several groups that were well-spread across the sky. SURVEY-6005 originally targeted members of BPMG, ABDMG, THA, COL, CAR, CARN, and UMA, however, updated group memberships have shifted SURVEY-6005 targets into a wider range of groups.

Table 3. Target summary showing for each assigned membership, the number of targets per spectral type, the total number of targets observed in the group, the age and uncertainties, and the median distance of our targets in that group, with uncertainties represented by the 16th/84th percentiles when more than one target is in the group. Memberships, spectral types, ages and distances for each target are obtained from MOCAdb.

MembershipAFGKMLTotalAgeDistance
Myrpc
ABDMG04380116133−10+15133^{+15}_{-10}37−13+1737^{+17}_{-13}
AUSC0001214342±318342 \pm31848−14+748^{+7}_{-14}
BPMG02143604326±326 \pm337−9+1437^{+14}_{-9}
CAR001300433−1+233^{+2}_{-1}57−4+457^{+4}_{-4}
CARN0012014200±50200 \pm5031−6+131^{+1}_{-6}
COL0330401033−1+233^{+2}_{-1}53−3+953^{+9}_{-3}
FCCN000100129±429 \pm446
FIELD001240711200±75011200 \pm75034−11+1834^{+18}_{-11}
MEL4000100124±124 \pm1109
OCTN000010155−25+4555^{+45}_{-25}52
SCOCEN000100115−15+1815^{+18}_{-15}136
THA06811302840±340 \pm346−4+946^{+9}_{-4}
TWA20031902410±310 \pm358−11+1658^{+16}_{-11}
UMA05960020414±23414 \pm2324−2+724^{+7}_{-2}
VUL0100001162±30162 \pm3032
TOTAL2212743693165

Table 3.

A complete breakdown of how many targets have been investigated for each program divided by spectral type is shown in Table 4. When divided by spectral types, L stars constitute roughly ∼2%\sim2\% of all the targets, M ∼42%\sim42\%, K ∼26%\sim26\%, G and F stars comprising about ∼16%\sim16\% and ∼13%\sim13\% respectively, and the rest are A type stars (∼1%\sim1\%).

Table 4. Target summary showing for each GO/SURVEY ID, the number of targets observed per spectral type and the total number of targets in the program.

IDAFGKMLTotal
4050200219023
5835001349356
600502126381086
TOTAL2212743693165

Table 4.

Finally, a complete description of each target is also presented in Appendix § 9.

Filter selection

NIRCam and MIRI coronagraphic simulations were employed using PanCAKE (an extension of the official JWST exposure time calculator; A. Carter [16]) to identify the most sensitive filters to sub-Jupiter mass companions (assuming clear and near solar metallicity atmospheric models from E. F. Linder et al. [46]). NIRCam filter F444W has been selected as the primary filter for this purpose, supported by the NIRCam F200W (for coronagraphy) and the F210M filter (for direct imaging; only for GO-5835), which provides the best mass sensitivity between the available short wavelength filters. One very important advantage coming from this set-up is that the combination of F200W+F444W (or F210M+F444W for direct imaging) filters enables a preliminary screening of our candidates based on their colors, since exoplanets are fainter in the F200W/F210M compared to F444W, whereas background stars will be brighter.

Observing strategy

We have performed NIRCam dual-band coronagraphic and imaging observations of a total of 165 targets (excluding references and misaligned). All coronagraphic observations have been taken using the MASK335R coronagraph mask (inner working angle: 0.82″5), in the SUB320A335R sub-array, while for the imaging observations we have used the SUBARRAY1_DITHER as the primary dither pattern with 4 Primary Dithers and the SUB160P sub-array. Each target exposure was set to approximately 30 minutes, and the specific readout parameters for any given observation were obtained and optimized using PanCAKE simulations ([16]).

While for GO-4050 we performed dedicated reference target observation, with 7 references taken, one for each imaging sequence, for GO-5835 and SURVEY-6005, we employed a self-referencing strategy to facilitate point spread function (PSF) subtraction through reference differential imaging (RDI - D. Lafrenière et al. [43]; G. Ruane et al. [72]). The goal behind the similar setup for all three surveys is to produce a homogeneous PSF library for all data, as will be explained in more detail in § 3.3.

Analysis

Data reduction

All data were reduced using the spaceKLIP6 pipeline (J. H. Girard et al. [35]; J. Kammerer et al. [38]; A. L. Carter et al. 2023). spaceKLIP is a Python-based data reduction and analysis toolkit for JWST high-contrast imaging and now also non-coronagraphic imaging data, designed to apply a specific sets of tools to transform “raw” uncalibrated data, to the final product suitable for science. Figure 1 summarize the major steps of the pipeline and their outputs for the case of a coronagraphic target. The full description for this process is presented in Ferrer-Chavez et al, submitted. The same steps have been followed for the imaging dataset as well excluding the ND squares masking, since they are not present in the imaging data.

Six-panel example of coronagraphic data reduction steps

Figure 1. Example of results from the data reduction steps for the coronagraphic dataset. Top row, from left to right: the original uncal data, as downloaded by MAST, detector calibrated calints data, and instrument calibrated data (after bad-pixels have been identified and removed by the pipeline). Bottom row, from left to right: pre-PSF subtraction shifted frame, additional padding has been added to the frame to allow for the shifting of the images; ND squares masked frame, to remove their impact on the sensitivity on overlapping areas; post-PSF subtraction residual. Now the sources are clearly visible in the derotated residuals, comprising one close-in (background) source not entirely visible before.

The raw JWST data (exposure raw data products designated by a file name suffix of “uncal” that usually contain only the raw detector pixel values from an exposure - left panel, top row, in Figure 1) are read into spaceKLIP and processed using the stages 1 and 2 from the jwst pipeline. During this phase, the pipeline account for dark subtraction, photometric calibration, saturation, non-linearity, ramp-fitting, background subtraction and flat-fielding, transforming the raw uncalibrated data into flux-calibrated cal/calints images (central panel, top row, in Figure 1). Once the images are calibrated, tools from the ImageTools module can be used to further process the cal/calints images and prepare them for PSF subtraction : i.e. to identify and clean bad pixels and apply median-subtraction - right panel, top row, in Figure 1; properly recenter and align all the images to a common center - left panel, bottom row, in Figure 1; mask the four Neutral Density (ND) squares present in NIRCam coronagraphic images - central panel, bottom row, in Figure 1. ND squares are used to perform target acquisition for NIRCam coronagraphy without saturating the detector. Because they heavily attenuate the flux in these regions, they must be masked before PSF subtraction to prevent artificial suppression during multi-image dither or rotation combinations.

RDI PSF subtraction

We performed PSF subtraction using the spaceKLIP implementation of pyKLIP7 (J. J. Wang et al. [83]), a python library for direct imaging of exoplanets and disks. pyKLIP uses an implementation of Karhunen-Loeve Image Processing (KLIP; R. Soummer et al. [75]) algorithms based off of principal component analysis to model and subtract off the stellar PSF, and KLIP-Forward Modeling (FM - L. Pueyo [70]) to perform the calibration of the contrast curves and companion fitting. Since pyKLIP requires a reference PSF library to perform RDI based PSF subtraction, for this project, we created two curated PSF libraries: one for the coronagraphic dataset and one for the imaging dataset that will be discussed in more detail in the next section 3.3. During this process, the science image can also be separated into a number of annuli and position-angle subsections. This allows us to perform the PSF subtraction independently in multiple subsections of the image, helping avoiding biases affecting the PSF subtraction quality.

An additional parameter to fine tune is the number of KL modes (i.e., principal components) used by the final PSF model for subtraction. One can either select a specific KL mode or allow spaceKLIP to use multiple modes, providing a cube of residual images after PSF subtraction, where each layer corresponds to removal of a different KL mode. This second approach has two major advantages. First it allows us to pick the best residual for each astronomical source. Since removing more modes (up to higher KL modes) is more aggressive at subtracting the PSF closer to the center of the star, having access to subtractions removing differing numbers of KL modes allows us to pick the best residual for each candidate, depending on its separation. Second, it allows us to detect false astronomical source that can be injected by reference stars in the science residual image. This usually happen when some of the reference stars used to build the PSF library have astrophysical sources in their field-of-view. These sources can end up injected in residual images after subtracting out KL modes that have inherited these sources, and usually appear and disappear between subtracting different KL modes, as the spurious sources are present in specific KL mode images. In contrast, a true astrophysical source will remain present as more and more KL modes are subtracted. These injected sources are usually first detected in these residuals and then tracked down and masked in their original frames when the final PSF-library is assembled for subtraction. For a more in depth analysis of the KLIP parameters that were adopted to reduce our dataset, we refer the reader to Ferrer-Chavez et al, submitted. The right panel, bottom row, of Figure 1 shows the typical outcome of this subtraction step, where candidate sources clearly appear in the residuals post KLIP subtraction.

Building A PSF Library for Robust RDI

For this project we assembled two curated PSF libraries built from sources drawn from the entire programs sample (i.e. self-referencing strategy): one for NIRCam coronagraphy, where we combined the F200W and F444W coronagraphic data from the three programs in a unique RDI PSF Library; and one for NIRCam imaging, where we only utilized F210M and F444W data from the GO-5835 program, since GO-4050 and SURVEY-6005 do not include any imaging targets.

Coronagraphic PSF library

All coronagraphic observations from programs GO-4050, GO-5835 and SURVEY-6005 are compiled into a single, large reference PSF library. Since the coronagraphic PSF library data reduction and performance characterization are laid out in more detail in a separate publication (Ferrer-Chavez et al. submitted), here we will provide the reader with a summary of the process.

A large PSF library provides useful PSF diversity (from variations in stellar spectral type, wavefront error realizations and precise coronagraph alignment) that can enhance the contrast achieved following PSF subtraction. However, non-PSF astrophysical signals present in the observations of the library (such as background stars or galaxies, close stellar binaries or bright debris disks) can contaminate the PSF subtracted image, in some cases even introducing false positives resembling planetary companions. For this reason, we visually inspect in detail all observations from the programs to remove from the PSF library any source of contamination. To do so, we first excluded all the observations with any contamination at close separations (within ∼3\sim3 arcseconds), and observations with exceedingly bright off-axis sources (such as bright foreground stars). These are contaminants that are bright enough to affect large portions of the image, and so they are difficult to properly mask. The remaining observations are included in a preliminary PSF library used to perform a preliminary PSF subtraction. The output of this process is then visually inspected to identify any remaining faint contaminants (see next section 3.3.2 for an example of this process). We clean these contaminants by masking a circular region centered at the source, and replacing those pixels with pixel-wise median across a set of contaminant-free stars similar to them (weighted by the flux of the star, to preserve PSF structure as much as possible).

Imaging PSF library

Similar to the coronagraphic case, to ensure that each science image could also serve as a reference PSF in the imaging PSF library (i.e., free of contaminating off-axis sources), we performed the same additional step of cleaning each reference of any contaminants.

Additionally, some specific targets have been completely removed from the list of imaging PSF references due to the presence of a bright off-axis companions that could not be efficiently masked, specifically: 2MASS-J04231498-1533245, 2MASS-J02495639-0557352, and 2MASS-J03550477-1032415. This left us with only 8 targets available as references in the imaging PSF library.

Identification of astrophysical objects in the final residual images Once the reduction and subtraction phases have been completed for all 165 scientific targets, the next step is to inspect each residual image for the presence of candidates sources in both the short and long wavelength images that may impact our sensitivity estimates. To be able to reliably inspect all the available data, we use an internal web application, built using Flask and utilizing js9 (a Javascript implementation of SAOimage ds9, [37]) that allows multiple users to visualize and interact with FITS images, and mark the positions of visually identified candidate sources.

As observations were taken we organized multiple campaigns where members of the team visually inspected and annotated sources present in the final residual images. To ensure sources are not duplicated based on the uncertainty in the by-hand marking process, any marked coordinates within 4 pixels of each other are median-combined to produce a single source position. This method has drawbacks for precise candidate identification, as it is certainly feasible for two sources to be within 4 pixels of each other, but is nevertheless sufficient for masking purposes. Due to the qualitative nature of visual inspection, it is possible that sources may have been missed by all annotators. However, with an average of ∼ 10 annotators for each image, it is unlikely that any sources bright enough to impact our understanding of the sensitivity limits have been missed. With the final coordinates of source objects as annotated by team-members, we mask the object position in the residual image using a circular mask whit a size depending on the brightness of the source. This insures that these objects do not affect our estimation of the sensitivity limits of the star around which they are found.

Results

Contrast curves

We constructed contrast curves to provide a preliminary measure of how well faint objects can be detected relative to the bright star under the coronagraph mask. spaceKLIP provides two separate metrics as part of its analysis tools package: the raw contrast curves and calibrated contrast curves.

To compute a raw contrast curve for each source, we first used spaceKLIP starphot.get_stellar_magnitudes tool that evaluates the source brightness and zero point flux for the relevant filter by fitting a scaled version of a stellar atmosphere model to available photometry8 ([86]). Then, the measured pixel brightness and the estimated source brightness are combined to determine the contrast at five times the standard deviation of the noise in a given annular region (5σ contrast), accounting also for small sample statistics ([55]). This threshold represents the faintest object that can be reliably detected above the noise.

Coronagraphic dataset reduction requires also accounting for mask throughput (i.e. the efficiency with which light passes through the coronagraph mask to the detector as a function of angular separation and wavelength J. E. Krist et al. [40, 41]). To correct for the spatial variations, spaceKLIP applies an instrumental throughput scaling factor to each location across the 2D image. At small separations, the mask significantly attenuates the host starlight but at the same time dims also any companion flux. This reduced throughput, paired with elevated residual stellar noise, steepens the contrast curve and degrades inner sensitivity limits. Conversely, at larger separations, the throughput approaches unity as mask suppression decreases, increasing sensitivity thresholds for faint companions. Because throughput also varies with wavelength, different filters produce distinct contrast curves depending on the coronagraph performance at those wavelengths.

While the raw contrast curves provide a quick, intuitive estimate of sensitivity, they do not fully reprodu duce the real performance of an imaging system because they do not account for post-processing effects that may impact the flux of a potential companion. To obtain a more accurate picture of the sensitivity of a system then, we need to compute a calibrated contrast curve that accounts for the effects of over-subtraction and self-subtraction of a companion during the PSF subtraction process. Throughput for high-contrast imaging algorithms like KLIP are typically low close to the star due to over-subtraction, where the algorithm aggressively removes residual starlight, and can unintentionally suppress a companion. At greater distances, the KLIP throughput improves, allowing the companion’s signal to remain largely unaffected by the subtraction process.

To compute the calibrated contrast curves, spaceKLIP performs injection and recovery of simulated companions into the data at known separations and position angles. To simulate the companion, the tool takes advantage of STPSF to generate the PSF. The algorithm then retrieves these injected signals and applies scale factors to correct for inaccuracies introduced during PSF subtraction. The peak flux of the injected companions is then measured in sigma relative to the 5σ5\sigma contrast at the injected separation. The retrieved scale factors are used to adjust the raw contrast curves, providing a more accurate representation of the true sensitivity achieved in a given observation.

Contrast Achieved on Isolated and Disk Targets

When analyzing the residuals, two major source of pollution are observed: a bright binary companion, and a disk. The heavy impact of binaries on F444W contrast sensitivity, is clearly shown in Figure 2, where the achieved calibrated contrast limit for binaries is almost an order of magnitude worst compared to either isolated stars or disks, which reach very similar contrast limits instead. On the other hand, analysis shows This shows that for our purposes, the impact that disks have on our sensitivity limits is mostly negligible in F444W (where the disk is barely present). Since our study is derived by the F444W sensitivity, the presence of the disk do not impact negatively our results, and they can safely be included in the final sample. Even in the filter F200W (where disks are usually brighter), analysis shows that the impact of the disks on contrast limits in this filter is mostly negligible as well.

Calibrated contrast curves for binaries, disks, and single sources

Figure 2. Calibrated contrast curve comparison between isolated sources (orange), single stars with disks (blue) and binaries without disks (magenta) in the GO-4050, GO-5835, and SURVEY-6005 surveys combined. To avoid confusion, the 4 targets with both binaries and disks were removed from this comparison. The plot shows the impact that binaries have on the contrast sensitivity compared to isolated source, or isolated sources hosting disks.

Binaries instead can not be included without an apriori careful treatment. In appendix § 10, we will present a new method developed for spaceKLIP to recenter and remove visually resolved binary companions tested on non-coronagraphic data, but since this will be the focus of an upcoming paper (e.g. Bruinsma et al. in prep.), they will they will be excluded by our current analysis.

For all further analysis then, we will focus only on the sensitivity limits of isolated targets and disks (with studies of specific disks being the focus of other papers in the series, e.g. K. A. Crotts et al. [24, 25]; S. Palatnick et al. [63] in the observed moving groups and associations). This selection leaves 111 isolated targets and disks for the sensitivity analysis.

Figure 3 shows the calibrated contrast curves for each program, where the thick line marks the median of the distribution, while the shaded area shows the 16th/84th16\mathrm{th}/84\mathrm{th} percentile for each distribution. The available field of view (FOV) coverage in the NIRCam coronagraphic (solid line) and imaging (dashed line) results in a lower achieved contrast at wide angular separations. For some specific datasets, small sample statistics have a major impact near the edge, causing the contrast to greatly deteriorate at wide separations. While the GO-4050 data roughly reach a minimum median contrast level of ∼10−5\sim10^{-5} at ≳2\gtrsim2 arcsec, the GO-5835 coronagraphic and SURVEY-6005 data are able to go much deeper in contrast (∼5×10−6\sim5 \times10^{-6} and ∼2×10−6\sim2 \times10^{-6} respectively). Since the targets in SURVEY-6005 are the only FGK targets in the surveys (excluding 1 K star in GO-4050 - see Table 4), the contrast curves for these sources are able to explore a much larger parameter space in contrast since the sources in this surveys are generally brighter compared to the M stars observed by GO-4050 and GO-

Calibrated contrast curves and magnitude limits for each program

Figure 3. Left: Calibrated contrast curves for each program. Right: Magnitude limits for each program. The thick continuous lines show the median of the distribution of curves, while the shaded areas cover the 16th/84th percentile for each distribution. The lower part of each plot shows the FOV coverage of the data (coronagraphic - solid line, imaging - dashed line) as a function of the distance from the center. The absolute magnitude sensitivity floor is governed by the instrument background noise rather than coronagraphic performance: GO-4050 and 5835 (mostly focused on M-dwarfs) exhibit poorer contrast while converging to identical magnitudes limits as SURVEY-6005 (primarily FGK hosts). For the GO-5835 imaging component, the absence of a coronagraph increased the overall throughput and signal-to-noise ratio achieved for this faint M-dwarf sample.

  1. Because all targets in the GO-5835 imaging survey are faint M-dwarfs (hence the decision to not use the coronagraph), the performance is the lowest, with a median contrast of only ∼6×10−5\sim6 \times10^{-5} at similar separations.

Magnitude limits

The right-hand panel of Figure 3 shows the magnitude sensitivity limits for each program. These curves are obtained by converting the contrast curve presented in Figure 3 into a magnitude difference, using the apparent magnitude of the corresponding target. To obtain the apparent magnitude of a target star, we used the starphot.get_stellar_magnitudes tool from spaceKLIP (see § 4.2).

Despite a significant spread in contrast sensitivity across the various coronagraphic programs, the median magnitude sensitivities consistently plateau at comparable limits. Specifically, SURVEY-6005 achieves a median F444W magnitude of ∼21.3\sim21.3 at separations ≳4\gtrsim4 arcsec, whereas GO-4050 and GO-5835 plateau at ∼21.4\sim21.4 at separations ≳2\gtrsim2 arcsec.

Ultimately, the absolute magnitude sensitivity floor is governed by the instrument background noise rather than coronagraphic performance. This background-limited regime explains why GO-4050 and 5835 (mostly focused on M-dwarfs) exhibit poorer contrast while converging to identical magnitudes limits as SURVEY-6005. Moreover, SURVEY-6005’s superior contrast does not improve its absolute magnitude sensitivity at small separations, primarily because its FGK hosts are intrinsically brighter than the M dwarfs in GO-4050/5835. Additionally, SURVEY-6005 exhibits marginally lower background sensitivity, as shorter readout ramps are required to prevent saturation, causing higher read noise compared to the longer DEEP8 ramps used in GO-4050/5835. The findings suggest that the necessary detector configuration for brighter targets limits the achievable sensitivity in the background-limited regime.

For the GO-5835 imaging component, the absence of a coronagraph increased the overall throughput and signal-to-noise ratio achieved for this faint M-dwarf sample. This yielded improved performance at closer separation, and a background magnitude limit in F444W that is ∼1\sim1 magnitude deeper than the coronagraphic limits (∼22.2\sim22.2 at ≳2\gtrsim2 arcsec).

Discussion

Survey sensitivity limits

To assess the sensitivity limits of the three JWST programs analyzed in this work, we converted the calibrated contrast curves presented in § 4.2, corrected for the finite field of view, into minimum detectable companion mass and effective-temperature detection limits.

To this end, we employed the Python tool MADYS9 [76], which derives statistical constraints on the detectability of self-luminous companions based on detection limits derived from high-contrast imaging observations. Suitable evolutionary models (described in § 5.2) are used to translate photometric contrast into companion masses and/or effective temperatures as a function of age.

Model selection

At the lowest temperatures expected to be probed by our programs (i.e. ≲350\lesssim350 K), exoplanet fluxes in the NIRCam bands will be at least partially affected by the presence of clouds [15, 79, 58, 11]. However, existing exoplanet atmospheric models (e.g., [73, 2, 47, 68, 53, 42, 59, 60, 71]) do not adequately cover the low temperatures required to evaluate the sensitivity we expect to reach, do not include clouds, or are uncoupled from evolutionary models. Additionally, recent JWST discoveries (such as TWA7b - [44, 24]) highlight the potential need for disequilibrium chemistry and enhanced metallicity to better characterize the flux of low-mass sub-Jupiter exoplanets, that current planet evolutionary models appears to overestimate. To compute sensitivity limits and detection probability maps, we then need to resort to 2 specific evolutionary model grids that cover sufficiently low temperatures for our purposes, combining the measured luminosity and the estimated age of each source to infer their mass and TeffT_{\mathrm{eff}}.

The first model, hereafter referred to as BEX–ATMO (or bex-atmo2023-ceq/neq-s/neq-w), combines the ATMO chemical-equilibrium atmospheric models ([69]) with the latest equation of state ([22]) and the BEX models ([46]), following the coupling scheme described in [20]. The second set of models is the Sonora Flame Skimmer models (or sonora-flame-skimmer-ceq/neq; [49 b]), the latest set of clear atmospheric and evolutionary models in the Sonora family ([53, 39, 59, 60, 85, 26]). These models span Teff=50T_{\mathrm{eff}} = 50 K to 2400 K, log⁡(g)=2−5.5\log(g) = 2 - 5.5 (cgs), a wide range of metallicity ranging from 10x sub-solar ([M/H] = -1.0) to 100x super-solar ([M/H] = +2.0), and C/O ratios from 0.5x solar to 2.5x solar (solar = 0.458) for both equilibrium (CEQ) and disequilibrium (NEQ) chemistry.

While all our models are currently cloud-free, the main difference between the CEQ models and the NEQ models is that the CEQ models allow all the chemical species to mix to reach a steady state with one another, while the NEQ models assume that vertical mixing dredges up different species into the upper layers of the atmosphere before they fully react with one another. Where these species mix is determined by when vertical mixing begins to dominate. This is parameterized by the eddy diffusion parameter, KzzK_{zz}, with various strengths of mixing ranging from Kzz=102,104,107,108K_{zz} = 10^2, 10^4, 10^7, 10^8, and 10910^9, where larger values of KzzK_{zz} represent stronger mixing. The parameters Fe/H and C/O instead regulate the composition of the objects (e.g. Jupiter has enhanced metallicity relative to the sun). The C/O, for example, is typically dictated by where in the disk the exoplanet formed relative to various ice/snow lines ([62]). The TeffT_{\mathrm{eff}} of the object also affects the dominant molecular species C and O are found in. For colder objects (late-T to Y dwarfs; ≲700\lesssim700 K), equilibrium chemistry favors O in H2_2O and C in CH4_4, while for hotter objects (L to early-T dwarfs; ≳1000\gtrsim1000 K) carbon is increasingly partitioned into CO and CO2_2. These temperature boundaries are approximate because the dominant reservoirs depend on the specific pressure-temperature profile of a given object ([52, 48, 87]).

Mass/Temperature limits

To study the sensitivity limits of our programs, we investigated models belonging to the BEX–ATMO (C/O = 0.55, Fe/H = 0.0) and Sonora Flame Skimmer families (C/O = [0.5,1,2.5], Fe/H = [0,1,2]). This range of parameters has been chosen to have a common point between Sonora and BEX–ATMO (C/O = 0.5 and Fe/H = 0) families. The Sonora models also offer Fe/H = −1, but since the candidate companions we expect to detect are likely to have solar or super-solar metallicity, we do not consider that case.

In Figure 4, we show a zoomed view of the mass/TeffT_{\mathrm{eff}} average sensitivity limits for each moving group. We limit the separation range plotted to [1, 4] arcsec, since this is the region where the majority of changes happen to our sensitivity, before plateauing to a constant value. We note that these results are dependent on the atmospheric model adopted and that we only plot moving groups with observations of N≥10N \ge10 members. Table 5 provide the mass/TeffT_{\mathrm{eff}} sensitivities at a separation of 2 arcsec for each moving group.

Mass/Teff sensitivity comparison between different groups

Figure 4. Mass/TeffT_{\rm eff} sensitivity comparison between different groups. Cloud-free BEX-ATMO and Sonora Flame Skimmer CEQ (continuous lines) and NEQ (dash-dotted lines), and a blackbody model (to act as a cloudy case scenario - green thick line) are shown for comparison. The yellow and blue horizontal dashed lines show the Jupiter and Saturn reference masses (top row), while the black horizontal dashed line shows the Epsilon Ind Ab reference temperature (bottom row - E. C. Matthews et al. [54]). Note that in the upper row, the moving groups have been ordered by their respective age (younger on the left). In the lower row instead, the moving groups have been ordered by their respective median distance (closer on the left), following the values reported in Table 3.

ModelCO/FeHTWABPMGCOLTHAABDMGUMAUMAABDMGBPMGTHACOLTWA
blackbody––––––383374371402415407
bex-atmo2023-ceq0.55/0.00.110.140.290.280.461.07233229230257268263
bex-atmo2023-neq-s0.55/0.00.110.140.290.280.461.06233229230257268263
bex-atmo2023-neq-w0.55/0.00.110.140.290.280.461.07234229230257268263
sonora-flame-skimmer-ceq0.5/0.00.100.130.300.290.471.20232231241261272272
0.5/1.00.110.150.280.280.461.11257254258277287283
0.5/2.00.090.130.300.300.461.13292290300325339351
1.0/0.00.100.130.290.280.471.17230230240259270272
1.0/1.00.110.150.270.270.451.10255253257276286282
1.0/2.00.100.140.270.250.421.05288289299321334347
2.5/0.00.100.120.290.280.461.15228228240258269271
2.5/1.00.110.150.270.270.451.09254252256275285282
2.5/2.00.140.170.370.360.731.34282283295315328341
sonora-flame-skimmer-neq0.5/0.00.110.180.360.360.521.29234236259285293305
0.5/1.00.270.450.740.721.122.69324315337356368378
0.5/2.00.240.400.670.671.302.53397390413449469471
1.0/0.00.130.180.480.490.591.41239243258299313286
1.0/1.00.280.510.820.781.292.79325327349367381393
1.0/2.00.280.450.720.721.122.81404390404441457470
2.5/0.00.140.260.490.470.721.72264273276294315299
2.5/1.00.370.631.161.141.643.43364349371412427420
2.5/2.00.410.711.221.221.994.29462441448494512510

Table 5. Sensitivity limits at 2.0 arcsec for Mass (MJupM_{\mathrm{Jup}}) and TeffT_{\mathrm{eff}} (K). Note that the moving groups have been arranged following the same order shown in Figure 4.

The top row of Figure 4 and the left side of Table 5 highlight the wide mass sensitivity range we will be able to reach with our programs. In TWA we can expect sensitivity to extend to exoplanets with sub-Saturn mass between ∼0.09−0.41\sim0.09 - 0.41 MJ_{\mathrm{J}}. In UMA instead, we are more likely to only reach super-Jupiter masses between ∼1.06−4.29\sim1.06 - 4.29 MJ_{\mathrm{J}} for enhanced metallicity. In both cases, the spread is dictated by the selected CEQ/NEQ models and the Fe/H and C/O parameters.

The bottom row of Figure 4 and the right side of Table 5 instead, show the TeffT_{\rm eff} versus separation space explored by our programs. As we stated at the beginning of § 2.2, at the temperatures probed by our programs we would expect clouds to be present in the atmosphere of the exoplanets. Similar to R. Bowens-Rubin et al. [11], we investigate the impact of cloud opacity on our TeffT_{\rm eff} sensitivity limits by estimating a maximally cloudy exoplanet as a blackbody. For each target, the F444W apparent magnitudes were computed as a function of the blackbody temperature and interpolated over the corresponding magnitude sensitivity curves (Figure 3). The resulting TeffT_{\rm eff}–separation curves were then averaged within each group to produce the final blackbody sensitivity limit (green thick lines) for the group. The bottom row of Figure 4 shows that while the presence of clouds can heavily impact our TeffT_{\rm eff} sensitivity limits compared to CEQ models, the impact of the Fe/H and C/O chemistry in NEQ models also have a strong effect on the suppression of flux received from exoplanets, pushing upward our overall TeffT_{\rm eff} sensitivity limits.

With all this considered, in UMA we could be sensitive to TeffT_{\rm eff} between ∼228−462\sim228 - 462 K for enhanced metallicity (the spreads again dictated by the amount of C/O). For TWA instead, we could be sensitive to Teff∼263−510T_{\rm eff} \sim263 - 510 K for exoplanets with enhanced metallicity.

To conclude this section, we provide in Figure 5, an overview of the sensitivity limits of our programs for the full separation range [0,∼14 arcsec][0, \sim14\ \mathrm{arcsec}]. This analysis, coupled with future atmospheric characterization, will be able to provide a better picture of what the true sensitivity limits of our programs are. Currently though, since there is not a real consensus on what the common atmospheric parameters for exoplanets in our programs should be, we decide to highlight what our worst/best-case constraints look like. For these specific reasons, we selected two models from the Sonora Flame Skimmer family, one for the CEQ case, with solar C/O and metallicity (C/O = 0.5 and Fe/H = 0.0), and one for the NEQ case, with more extreme parameters (C/O = 2.5 and Fe/H = 2.0), since they allow for a greater portion of parameter space to be explored compare to all the other model parameters, and the BEX-ATMO model as well. Similar to Figure 4, we grouped together the different sources based on their groups, but in this case we overplotted one on to the other for a quick comparison. The assumed age and median distance of each group (as reported in Table 3) are displayed for reference. Figure 5 clearly shows that past ∼2\sim2”, Jupiter masses (or Saturn masses in the most optimistic case) and Teff≲275T_{\mathrm{eff}} \lesssim275 are well within reach of our programs.

Mass and effective-temperature sensitivity limits for the Sonora Flame Skimmer family models

Figure 5. Mass (top)/TeffT_{\mathrm{eff}} (bottom) sensitivity limits for the Sonora Flame Skimmer family models with C/O = 0.5, Fe/H = 0.0, CEQ (left); and C/O = 2.5, Fe/H = 2.0, NEQ (right). The yellow and blue horizontal dashed lines show the Jupiter and Saturn reference masses, while the black horizontal dashed line shows the Epsilon Ind Ab reference temperature (E. C. Matthews et al. [54]). The shaded area cover the 16th/84th percentile for each distribution.

Temperature sensitivity limits comparison between NIRCAM and MIRI

Similar to R. Bowens-Rubin et al. [11], we perform a comparison between the expected TeffT_{\mathrm{eff}} sensitivities for both JWST MIRI F2100W imaging and NIRCam F444W coronagraphy, based on our ample sample. In their analysis, they compared equilibrium models for three atmospheric cases: clear, cloudy (fsed=8f_{\mathrm{sed}} = 8), and maximally cloudy, concluding that MIRI imaging offers the advantage for systems within 20 pc. In our analysis instead, we explored NIRCam performances for both equilibrium and non-equilibrium cloud-free models, for different values of C/O and Fe/H, and then compared against their predictions. A blackbody model is also included to act as a cloudy case comparison scenario. To facilitate a direct comparison to their NIRCam/MIRI performance, we converted their 5-σ\sigma magnitude limits in the background-limited regime for EV Lac (20.2 mag in F444W, 15.1 mag in F2100W, at 2.6 arcsec) to the expected TeffT_{\mathrm{eff}} in the Sonora Flame Skimmer models with C/O=[0.5,2.5] and Fe/H=[0,2], after rescaling them to different distances to trace how their estimated performances would look like in our reference system.

In Figure 6 we present the predicted value of TeffT_{\mathrm{eff}} sensitivity at 2 arcsec NIRCam and MIRI instruments. Similar to Figure 4, a blackbody model (green thick line) is overplotted to both panels to act as a cloudy case comparison scenario. Continuous color code lines shows the different model parameters. The red (F444W) or blue (F2100W) lines shows R. Bowens-Rubin et al. [11] converted 5-σ\sigma magnitude limits in the background-limited regime for the C/O = 0.5, Fe/H = 0 (dashed line), and C/O = 2.5, Fe/H = 2.0 (dotted line) case. The triangles represent their converted 5-σ\sigma background-limited estimation of EV Lac, from where the extrapolation to higher distances begun. The Figure shows (as expected) how the minimum TeffT_{\mathrm{eff}} sensitivity decrease as farther the group is. Distant sources require higher TeffT_{\mathrm{eff}} to offset distance-related flux attenuation (F∝Teff4/d2F \propto T_{\mathrm{eff}}^4/d^2) and remain above the instrumental detection threshold. Also in this case we see an outlier in UMA (∼25\sim25 pc), that seems to perform much worse compared to the rest of the trend, or TWA compared to COL, that performs better. One possible reason is that while we used the age of the cluster as reference age for all the members of the same group, for the distance we used individual reported values for each target, leaving us with a wider distribution of distances in each group (sometimes overlapping

Effective-temperature sensitivity at 2 arcsecond for the CEQ and NEQ models

Figure 6. TeffT_{\mathrm{eff}} sensitivity at 2 arcsecond, for the CEQ (left) and NEQ models (right), showing an overall good agreement between our sensitivity limits and R. Bowens-Rubin et al. [11] predictions, even though the pace at which MIRI can potentially outperform NIRCam TeffT_{\mathrm{eff}} sensitivity depends on atmospheric chemistry of the expected population of exoplanets. Each dot represent the TeffT_{\mathrm{eff}} sensitivity for a different group: UMA, ABDMG, BPMG, THA, COL and TWA. Each continuous line instead shows either the BEX-ATMO (black), or the Sonora Flame Skimmer (color coded by their different parameters) models. A blackbody model is overplotted to act as a cloudy case comparison (green thick line). 5-σ\sigma sensitivity from R. Bowens-Rubin et al. [11] for F444W (blue) and F2100W (red), converted into our Sonora Flame Skimmer models, are overplotted for the [C/O, Fe/H] = [0.5, 0.0] (dashed line), and [C/O, Fe/H] = [2.5, 2.0] (dotted line) case. The triangles represent the original point from where the extrapolation to higher distances begun.

with in 1σ1\sigma), that could lead to much similar sensitivity differences when the median is taken. Additionally, the spectral type composition of each group plays a role in the sensitivity limits. While the UMA sample is mostly composed by F and G stars, the BPMG sample, the next group in distance, is mostly composed of M-dwarfs. The same reasoning goes for the comparison between COL and TWA samples, where COL has a more balanced spectral type composition, with slightly more F, G stars compared to K or M, while TWA sample is mostly composed of M-dwarfs. As we have seen from Figure 8, the spectral composition has an impact on the performances the TeffT_{\mathrm{eff}} sensitivities. Overall, our sensitivities mostly match R. Bowens-Rubin et al. [11] predictions, with a particular good agreement in the CEQ plot. When we compare to their MIRI results instead, we see that our NIRCam sensitivity limits start to out pace their MIRI predictions at ∼25−30\sim25-30 pc (best/worse case scenario) for the Sonora Flame Skimmer CEQ models, and at ∼25−50\sim25-50 pc (best/worse case scenario) for the Sonora Flame Skimmer NEQ models.

Mass/$T_{\mathrm{eff}}$ detection probability maps comparison for two Sonora Flame Skimmer CEQ and NEQ models

Figure 8. Mass/TeffT_{\mathrm{eff}} detection probability maps comparison for two Sonora Flame Skimmer CEQ and NEQ models, for the A, F, G (combined), and K and M (combined) spectral types. All maps are derived from the F444W sensitivity limits shown in Figure 3. The yellow and blue horizontal dashed lines show the Jupiter and Saturn reference masses, while the black horizontal dashed line shows the Epsilon Ind AB reference temperature (E. C. Matthews et al. [54]). The last row in both panels clearly show not only the considered atmospheric chemistry, but also the spectral type of the targeted star has an heavy impact on the expected planet Mass/TeffT_{\mathrm{eff}} detection probabilities.

Finally, a comparison with our cloudy case (the black body line) shows that while clouds may have a suppressive effect for NIRCam coronagraphic observations (as shown by the CEQ panel of Figure 6, as well as R. Bowens-Rubin et al. [11]), increasing metallicity might have a greater potential to limit F444W observations (see NEQ panel of Figure 6). Since both clouds fractions and metallicity are quite uncertain for this class of objects, only future atmospheric characterization of the exoplanet population uncovered by GO-4050, GO-5835, and SURVEY-6005 (i.e. the upcoming analysis from GO-10764 program) will be able to disentangle these two cases. This results, based on the analysis of 141 targets (the sum of all the UMA, ABDMG, BPMG, THA, COL and TWA target members), while mostly confirming R. Bowens-Rubin et al. [11] predictions, also highlight once again the tension between the atmospheric chemistry of the expected population of exoplanets and the instrument predicted sensitivity limits.

Detection probability maps

To be able to place population based general mass and TeffT_{\mathrm{eff}} predictions, we used MADYS (see § 5.1) to convert our calibrated contrast curves into detection probability maps, using the same models we introduced earlier in § 5.2. The detection probability maps are computed over a grid in semi-major axis and companion mass spanning the range of [1,1000][1, 1000] au ×[0.005,10]\times[0.005, 10] MJM_{\mathrm{J}}, uniformly sampled in logarithmic space. At each grid point, N=1000N = 1000 sets of orbital elements are randomly drawn and used to convert the semi-major axis into projected separations; in particular, inclinations are sampled from a distribution uniform in cos⁡i\cos i, while eccentricities are drawn from a truncated normal distribution N(0,0.9)\mathcal{N}(0, 0.9). In parallel, the stellar age constraint is sampled to convert the companion mass into a set of apparent magnitudes in the relevant filter.

For each realization, a simulated companion is considered detected if its contrast at the corresponding projected separation exceeds the 5σ5\sigma contrast limit. Averaging over all realizations yields a mean detection probability that self-consistently accounts for projection effects and age uncertainties. Detection probability maps in effective temperature are then obtained by remapping the mass–semi-major axis probability grid using the model-predicted mass–temperature relation at each sampled age. Similar to the end of the precedent section, in this section we will also limit our analysis to two specific Sonora Flame Skimmer models, with a standard solar value for C/O =0.5= 0.5 and Fe/H =0.0= 0.0 for the CEQ case, and an enhanced extreme case of C/O =2.5= 2.5 and Fe/H =2.0= 2.0 for the NEQ case, as references for our discussion.

In the following sections, we will present our detection probability maps as follows. The first two rows show the results for the selected model, and for each selected category (left and central columns). The right column instead shows the difference between the first category (blue) and the second (red). We will refer to this direction along each row as “model-wise”. The left and central panels of the bottom row instead show the difference between the specific category in the sonora-flame-skimmer-ceq model (green) and the sonora-flame-skimmer-neq model (yellow). We will refer to this direction along each column as “category-wise”.

To offer a fairer comparison (since we can not show the entire distribution of performances, e.g. as in Figure 5), these plots are created taking into account equal-size samples. To do so, for each category we ranked the detection probability maps based on the sum of their detection probability below 11 MJM_{\mathrm{J}} (or 10001000 KK). Then we picked the best maps so that both categories have the same number of targets, and average them. Detection probability maps for the most best-sampled groups in our programs (TWA, BPMG, COL, THA, ABDMG, and UMA) are shown in appendix § 11.

Filter comparison

Figure 7 shows the comparison between detection probability maps as a function of mass/TeffT_{\mathrm{eff}} versus separation for the filters F200W and F444W coronagraphic data. Note that for this comparison, we have to exclude into projected separations; in particular, inclinations are sampled from a distribution uniform in cos⁡i\cos i, while eccentricities are drawn from a truncated normal distribution N(0,0.9)\mathcal{N}(0, 0.9). In parallel, the stellar age constraint is sampled to convert the companion mass into a set of apparent magnitudes in the relevant filter.

Mass/Teff detection probability maps comparison for two Sonora Flame Skimmer CEQ and NEQ models

Figure 7. Mass/TeffT_{\mathrm{eff}} detection probability maps comparison for two Sonora Flame Skimmer CEQ and NEQ models, for filters F200W and F444W. The yellow and blue horizontal dashed lines show the Jupiter and Saturn reference masses, while the black horizontal dashed line shows the Epsilon Ind AB reference temperature ([54]). The last row in both panels clearly show the impact of atmospheric chemistry (e.g. disequilibrium chemistry, enhanced metallicity) on the planet Mass/TeffT_{\mathrm{eff}} detection probabilities.

For each realization, a simulated companion is considered detected if its contrast at the corresponding projected separation exceeds the 5σ5\sigma contrast limit. Averaging over all realizations yields a mean detection probability that self-consistently accounts for projection effects and age uncertainties. Detection probability maps in effective temperature are then obtained by remapping the mass–semi-major axis probability grid using the model-predicted mass–temperature relation at each sampled age. Similar to the end of the precedent section, in this section we will also limit our analysis to two specific Sonora Flame Skimmer models, with a standard solar value for C/O =0.5= 0.5 and Fe/H =0.0= 0.0 for the CEQ case, and an enhanced extreme case of C/O =2.5= 2.5 and Fe/H =2.0= 2.0 for the NEQ case, as references for our discussion.

For each realization, a simulated companion is considered detected if its contrast at the corresponding projected separation exceeds the 5σ5\sigma contrast limit. Averaging over all realizations yields a mean detection probability that self-consistently accounts for projection effects and age uncertainties. Detection probability maps in effective temperature are then obtained by remapping the mass–semi-major axis probability grid using the model-predicted mass–temperature relation at each sampled age. Similar to the end of the precedent section, in this section we will also limit our analysis to two specific Sonora Flame Skimmer models, with a standard solar value for C/O =0.5= 0.5 and Fe/H =0.0= 0.0 for the CEQ case, and an enhanced extreme case of C/O =2.5= 2.5 and Fe/H =2.0= 2.0 for the NEQ case, as references for our discussion.

the non-coronagraphic sources in the GO-5835 dataset, since they do not have F200W measurement, lowering the total number of targets to 103.

Model-wise, the F444W filter consistently outperforms the bluer F200W filter in mass and TeffT_{\mathrm{eff}} sensitivity. Because the SEDs of cool (≲500\lesssim500 K) planetary-mass companions peak in F444W while stellar flux declines, this bandpass provides the primary detection layer for identifying candidates. Overall, the sonora-flame-skimmer-ceq models reach a peak of their difference at ∼75%\sim75\% detection probability around ∼1\sim1 MJ_{\mathrm{J}} or ∼400\sim400 TeffT_{\mathrm{eff}} at ∼200\sim200 au, while the sonora-flame-skimmer-neq models barely reaches ∼25%\sim25\% for ∼1\sim1 MJ_{\mathrm{J}}, and ∼50%\sim50\% for ∼500\sim500 TeffT_{\mathrm{eff}}. This dichotomy will have a direct impact on the final yield achieved by our three programs, and in turn, will also better inform us on the atmospheric chemistry of the exoplanet population (see also later in § 5.7).

Conversely, a filter-wise comparison reveals that F200W detection probabilities remain uniform across the models. In contrast, the F444W filter is highly sensitive to the choice of equilibrium versus disequilibrium chemistry and model parameters. For separations between ∼50\sim50 and 500500 au, the model-to-model variation in F444W detection probability peaks at ∼50%\sim50\% for masses between (≲1\lesssim1 MJ_{\mathrm{J}}) and ∼1\sim1 MSat_{\mathrm{Sat}}, and reaches ∼75%\sim75\% for effective temperatures of ∼300\sim300–400400 K.

This result strongly suggests that atmospheric chemistry (e.g. disequilibrium chemistry, enhanced metallicity) of the planets we are considering can then heavily affect our detection thresholds (especially at the lower mass/TeffT_{\mathrm{eff}} limits). Then, the potential Neptune- to Saturn-mass planet inferred from cold debris disc structures at 10s of au ([65]), might be currently just inside our detectability window, depending on their own chemical composition.

Spectral type comparison

Figure 8 shows a comparison between detection probability maps as a function of mass/TeffT_{\mathrm{eff}} versus separation for different spectral types for the F444W filter. Note that for this comparison instead, we excluded the three L-dwarfs, lowering the total number of targets to 108. We then split these 108 targets in two samples: A, F, and G spectral types stars (37 targets in total), and K and M spectral types stars (71 targets in total). Finally, since we want to perform an equal-size comparison (as explained at the end of § 5.5) we selected the 37 best performing targets from the K and M sample, ending up with equal number of targets for each sample. Similar to Figure 7, we observe an increase in sensitivity towards the lowest masses for later spectral types. Model-wise, the K and M spectral types sample outperform the A, F, and G, with a detection probability ∼50%\sim50\% in favor of the later spectral types for ≲1\lesssim1 MJ_{\mathrm{J}} already 20 au, and ∼75%\sim75\% for ≲1\lesssim1 MSaturn_{\mathrm{Saturn}} at 60 au for the sonora-flame-skimmer-ceq. The sonora-flame-skimmer-neq models instead shows a shift upward of the detection probability (very clear when looking at the plots spectral-type-wise, or SpT-wise for short), with a detection probability ∼75%\sim75\% in favor of the K and M group for ∼1\sim1 MJ_{\mathrm{J}} at 100 au, and just a few percent for ∼1\sim1 MSaturn_{\mathrm{Saturn}}. When looking at the TeffT_{\mathrm{eff}} instead, the sonora-flame-skimmer-ceq reach a peak of the detection probability at ∼50%\sim50\% at ∼350\sim350 K at ∼30\sim30 au, while the sonora-flame-skimmer-neq models reach similar values at ∼600\sim600 K and ∼40\sim40 au.

This dependence with the atmospheric chemistry of the exoplanets is made even clear in the last row of both panels in Figure 8, where we show the SpT-wise variation across the two models. For the mass detection probability maps (top panel), the sonora-flame-skimmer-ceq model provides an increase between ∼25\sim25–50%50\% on the detection probability around A, F, and G stars, between one Saturn- to Jupiter-mass at ∼30\sim30–300300 au. For the K and M stars instead, a dramatic shift appears, with the detection probability reaching up to ∼75%\sim75\% between sub Saturn (∼0.1\sim0.1 MJ_{\mathrm{J}}) - to Jupiter-mass at similar separations (in agreement with predictions from [16]). The TeffT_{\mathrm{eff}} sensitivity instead (bottom panel) appears to be more balanced between the different spectral types, with both the A, F, and G sample, and K and M sample, reaching ∼75%\sim75\% more the detection probability.

We need to notice at this point that, when building these maps, targets with different ages and distances are included in each sample, and this mixing can bias the final detection probability map for each spectral type. The mass sensitivity for example will be more sensitive to shifts in ages compared to the TeffT_{\mathrm{eff}} sensitivity. Thus, it is not easy to disentangle from this approach how much of the differences we observe is due to the different spectral types, than from the fact that the F, G, and K star sample has a higher average age versus the M stars (see Table 3 for reference). To address this problem, we present in the next § 5.6.1 a principal component analysis of the detection probability maps, with the goal of disentangling the contribution of the spectral type, age and distance of the targets on the detection probability maps.

Principal Component Analysis of Detection Probability Maps

The analysis presented in § 5.6 shows how degenerate the problem of interpreting variations in the detection probability maps is with the spectral type, age, and distance of the sources in the maps. This inherently complex situation stems from the heterogeneity in the design of the constituent JWST programs (GO-4050, GO-5835, and SURVEY-6005). GO-4050 and GO-5835 observed primarily lower-mass M star targets in TWA and BPMG, with ages of either ∼10\sim10 Myr or ∼22\sim22 Myr, whereas SURVEY-6005 observed a sample of solar analog FGK stars, distributed among somewhat older groups (∼22\sim22 Myr for the youngest BPMG members in the sample, ∼400\sim400 Myr for the oldest UMA members in the sample). Thus, given the inherent inhomogeneity of the sample, it is not entirely possible to disentangle the effects of age, distance, and spectral type.

However this configuration offers also a unique opportunity: an unprecedented large, self consistent sample of hundreds of targets with consistent observations in the same observing mode from multiple moving group and associations, covering a wide range of spectral types, ages and distances. To more unambiguously determine the dominant variable(s) in producing the observed mass sensitivity of the sub-Jupiter survey observations, we performed a principal component analysis (PCA) on the set of individual target mass detection probability maps across the full sub-Jupiters survey sample. We selected only single targets with no evidence of a disk in F200W imaging. For this sample of 93 stars, we used code adapted from the pyKLIP pipeline [83] to build a Karhunen–Loève (KL) eigenbasis from the detection probability maps of these stars.

pyKLIP is a PCA pipeline designed specifically for high-contrast imaging; in its more usual usage, it builds a reference library of high-contrast images (either a series of images taken over time at different position angles or a series of different reference star images), builds a KL eigenbasis from this library, and using this basis, reconstructs a target image from a selected number of modes from the KL eigenbasis [75, 3]. The reconstructed image is then subtracted from the original target image; as any astrophysical companion is either not present in the reference library (in the case of a reference library built from observations of other stars) or is moving in position angle in the reference library (in the case of a reference library built from a time series of objects taken at different position angles), this removes light from the star and speckles, while retaining light from any astrophysical object in the field.

In our adaptation of this code to the sub-Jupiter survey detection probability maps, we build a KL eigenbasis from the full set of detection probability maps, to determine how many of the constructed eigenimages are necessary to describe most of the variance in the dataset. PCA is a dimensional reduction technique; by construction, the first eigenimage / principal component (henceforth PC) explains most of the variance in the dataset, followed by each successive PC. We find that the first PC describes 83.5% of the total variance, while the second PC describes 7.7% of the total variance, and the third PC describes 4.2% of the total variance. The remaining PCs account for only 4.7% of the total variance. Thus, In this case, one PC / eigenimage describes the majority of the variance in the system.

We then projected each of the detection probability map images onto the constructed KL eigenbasis, in order to associate each PC with the true physical parameters determining detection probability map sensitivity: stellar mass (used here as proxy for the spectral type), stellar age, and system distance. In Figure 9, we plot the PC scores (the ”weight” of the projection of each image onto each of the PCs) for the first two principal components. In the left panel, the color of the data points reflect the age of the star, whereas the color of the data points reflects the stellar mass in the middle panel, and the distance to the system in the right panel. These PC scores are not meaningful on their own, but can provide a sense of which of the physical parameters of the system are connected to each of the PCs. The PC scores appears to be clearly correlated with system age, while seem to be more weakly correlated with stellar mass, and no clear connection is seen with distance. Thus, this analysis shows that system age is the most important variable in determining the achieved detection probability map for a given star from our sample.

Figure 9. PC scores for the first two principal components, which account for more than 90% of the total variance in the sub-Jupiters DPMs of single stars without disks. The PC scores appears to be clearly correlated with the system age (left), while weakly correlating with the stellar mass (center). No clear connection is observed with the distance (right).

Comparison with PanCAKE simulations

In our original survey design, we used PanCAKE (an extension of the official JWST exposure time calculator; [16]) to simulate observations of survey targets in order identify the most sensitive filters for this project. In this section, we discuss the comparison between the actual achieved detection probability performance in our programs in filter F444W and the predicted result from the corresponding PanCAKE10 simulations for 97 of our coronagraphic targets. Figure 10 illustrates the comparison between PanCAKE simulations and our empirical data. Both panels demonstrate strong agreement across all modes, masses, and effective temperatures (TeffT_{\rm eff}). The PanCAKE simulations accurately predict NIRCam coronagraphic performance, with discrepancies remaining well below ∼10%\sim10\% at separations ≳30\gtrsim30 au. For closer separation, our sensitivities outperform PanCAKE prediction, with a discrepancy of ∼25%\sim25\%. The likely explanation is that the employed PSF library subtraction provides improved performance over what was assumed for PanCAKE (a 9-point single reference dither). This is supported by the fact that the improvement is limited to the closest separations (≲30\lesssim30 au), where the sensitivity limit is in a contrast-limited regime. Moreover, these results appear to be in contrast with predictions obtained for MIRI, where PanCAKE usually tends to under predict the instrument performance ([8]).

Mass/Teff detection probability maps comparison

Figure 10. Mass/TeffT_{\mathrm{eff}} detection probability maps comparison for two Sonora Flame Skimmer models as explained in the text, between the PanCAKE simulations of our targets and our programs in F444W. The yellow/blue dashed lines show the Jupiter and Saturn reference mass and TeffT_{\mathrm{eff}}. All comparisons are based on equal size samples, as explained in the text.

Comparison with Bogat et al. 2025

Figure 11 illustrates the comparison between our work and the GTO-1184 survey (E. Bogat et al. [10]). E. Bogat et al. [10] obtained F356M and F444W NIRCam coronagraphic imaging of nine nearby (≲21.1\lesssim21.1 pc) M-dwarfs, spanning a range of ages similar to our targets’ ages (i.e., ∼24\sim24–440440 Myr). Note that to build this comparison we have to exclude non-coronagraphic sources in the GO-5835 dataset, lowering the initial number of targets in our combine programs to 103.

Mass/Teff detection probability maps comparison for two Sonora Flame Skimmer models

Figure 11. Mass/TeffT_{\rm eff} detection probability maps comparison for two Sonora Flame Skimmer models as explained in the text, between GTO-1184 and our programs in F444W. The yellow/blue dashed lines show the Jupiter and Saturn reference mass and TeffT_{\rm eff}. All comparisons are based on equal size samples, as explained in the text.

Since our two sample are very different in size, we decide to pick the best 9 performing targets as our reference sample for this comparison (as explained at the end of § 5.5). This process left us with 9 M-dwarfs selected from our initial sample, with distances between ∼34\sim34–6363 pc, and ages ∼10\sim10 Myr.

When compared to our selected sample, we see that GTO-1184 greatly overperforms at closer separations (≲10\lesssim10 au), and overall is able to potentially reach both lower detectable masses and TeffT_{\rm eff}. This result clearly highlight the cardinal role that also the distance plays when trying to detect close, low-mass, cool sources.

Comparison with the SPHERE SHINE survey

The VLT SpHere INfrared survey for Exoplanets (SHINE; [27, 45, 82, 23]) obtained near-IR (11-22 μm) high contrast imaging of 400 nearby, young stars using the VLT/SPHERE (Spectro-Polarimetric High-contrast Exoplanet REsearch) IRDIS dual-band imager. The full SHINE sample and data reduction/processing is presented in [23]; [78]. Similar to our survey sample, the SHINE sample is largely drawn from nearby young moving groups and young stellar associations, but also includes other stars with indications of youth in the solar neighborhood ([27]). The SPHERE-SHINE sample covers a wider range of spectral types than our own survey, with about a third of this sample comprising of higher mass AB stars and with only 20 M stars included.

In Figure 12, we present a comparison between the mass sensitivities of the SHINE survey and the JWST NIRCam survey. As previously, we use Sonora Flame Skimmer models to convert from contrast sensitivity to mass sensitivity for JWST observations; however, the SHINE survey uses the BEX-COND models for this purpose ([1, 51]). Given the differing coverage of spectral types between the two surveys, to ensure a robust comparison, for each spectral type in our programs (i.e. A, F, G, K, M, and L), we constructed equal-size sub-samples from the SHINE survey (see previous section 5.5) ensuring that the target count as a function of spectral type in this comparison precisely matches between the two surveys compared. At the end of this process, we ended up with a final sample of 106 targets for each category. While this selection approach guarantees matching sample sizes and spectral type distributions, the individual targets remain distinct across the two surveys. Furthermore, because different evolutionary models were utilized to derive companion masses, direct comparisons

Mass detection probability maps comparing SHINE and F444W observations

Figure 12. Mass detection probability maps comparison for the two Sonora Flame Skimmer models as explained in the text, between a matched sample across the SHINE survey and our programs in F444W. Note that for the SHINE survey, the BEX-COND models have been used instead of the one showed by the y-labels. All comparisons are based on equal size samples, as explained in the text. The yellow/blue dashed lines show the Jupiter and Saturn reference mass.

between their mean detection probabilities cannot be interpreted as strictly one-to-one.

The SHINE survey is the most sensitive surveys for directly imaged exoplanets from the ground to date. However, the improvements in mass sensitivity provided by the extreme sensitivity of JWST compared to ground-based instruments provides access to a completely new parameter space for exoplanets, yielding sensitivity to sub-Jupiter exoplanets at wide separations from their stars. While the SHINE survey has always used the BEX-COND models (F. Allard et al. [1]; G.-D. Marleau et al. [51]), for the JWST programs in this case we alternate between the sonora-flame-skimmer-ceq model (our best case scenario - first row) and the sonora-flame-skimmer-neq model (our worst case scenario - second row). In the first case we are able to reach similar detection probability compared to ground observations already at ∼10\sim10 au and ∼3 MJ\sim3\,M_{\mathrm{J}}, and ∼50%\sim50\% in mass sensitivity between ∼50−500\sim50-500 au in the region between ∼1 MJ\sim1\,M_{\mathrm{J}} to ∼1 MSat\sim1\,M_{\mathrm{Sat}}. Since the stellar age distribution sampled by the SHINE survey is consistent with our Programs (e.g. ∼90%\sim90\% of the stars in the SHINE survey have an estimated age ≲450\lesssim450 Myr; V. Squicciarini et al. [78]), the impact of individual age discrepancies on the average sensitivity is expected to be limited.

When instead we compare to our more extreme case ([C/O,Fe/H]=[2.5,2.0][\mathrm{C/O}, \mathrm{Fe/H}] = [2.5, 2.0]), we see an inversion of this trend, with the SHINE survey outperforming the JWST observation at lower masses, at almost any separations. This latest comparison though, represents more of an upper bound to our sensitivities, due to the extreme nature of the high metallicity model. Indeed, the detection of TWA-7b (K. A. Crotts et al. [24]) in our sample already provides proof that we are actually not in this worst-case scenario presented here, and that JWST does have significant advantages over SPHERE.

Exoplanet yield estimates

While survey observations are not yet complete and thus the true yield of our survey is not yet known, here we make predictions of the expected number of new exoplanet discoveries to come from this work, given different models of the underlying exoplanet distribution.

Using core accretion synthetic population models

We restrict this analysis to single FGKM stars belonging to young moving groups and lacking disk detections, yielding a sample of 97 stars. To estimate the number of companions our JWST survey might detect, we divided the observed sample into two stellar-mass bins, separating Sun-like FG stars from lower-mass KM dwarfs, which are expected to host different planetary populations. We used the “New Generation Planetary Population Synthesis” (NGPPS) models for the accretion of planetesimals (A. Emsenhuber et al. [28]; R. Burn et al. [14]), adopting models for 1 M⊙1\,M_{\odot} and 0.7 M⊙0.7\,M_{\odot} primaries as proxies for the underlying core accretion planet populations around FG and KM hosts. By drawing FG and KM populations separately from models that best match their respective host masses, the resulting synthetic sample better reflects the stellar mass distribution of our survey.

These synthetic populations were normalized in three complementary ways: using the intrinsic occurrence predictions of the population synthesis models themselves, the SPHERE/SHINE direct imaging constraints at wide separations, radial velocity occurrence rates at tighter separations. The empirical normalizations assign a scaling factor to original synthetic populations such that the number of companions falling within a chosen calibration region reproduces the measured occurrence rates (see left panel of Figure 13). For the SHINE survey normalization, we used the occurrence constraints derived over 11–75 MJup75\,M_{\mathrm{Jup}} and 55–500 au500\,\mathrm{au} as part of the final SPHERE GTO statistical analysis (Chauvin et al. in prep.), adopting the core accretion occurrences rates measured for 0.80.8–1.5 M⊙1.5\,M_{\odot} and <0.8 M⊙<0.8\,M_{\odot} hosts using the same synthetic population models as those employed here. For the radial velocity normalization, we used the California Legacy Survey results for giant planets between 22–8 au8\,\mathrm{au} around Sun-like stars (B. J. Fulton et al. [32]), and HARPS occurrence rates for giant planets with periods of 1,0001{,}000–10,00010{,}000 days (∼1.5\sim1.5–7 au7\,\mathrm{au}) around M dwarfs (L. Mignon et al. [57]).

Combined synthetic populations and expected survey yield

Figure 13. Left: Combined 1 M⊙M_{\odot} and 0.7 M⊙M_{\odot} NGPPS core accretion synthetic populations, each sampled 20 times with a 0.2 dex log-space perturbation in mass and semi-major axis to improve sampling at wide separations. The golden colormap shows the average detection completeness for the 97 stars from the survey, for companions up to 10 MJupM_{\mathrm{Jup}}, for the sonora-flame-skimmer-ceq model with C/O = 0.5 and [Fe/H] = 0.0, with contours marking the 20%, 50%, and 80% detection probability levels. The teal and purple boxes indicate the regions of parameter space used to normalize the 1 M⊙M_{\odot} and 0.7 M⊙M_{\odot} NGPPS synthetic populations to the SHINE direct imaging and radial velocity occurrence constraints, respectively. Right: distributions of the expected survey yield under the three normalization cases: the intrinsic NGPPS model frequencies (grey), SHINE occurrence rates (teal), and RV measurements (purple), for the same atmospheric predictions.

To predict the survey yield under each normalization and evolutionary model, we followed a Monte-Carlo procedure. For each realization, we drew two stellar subsamples from the normalized synthetic populations: NFGN_{\mathrm{FG}} stars from the 1 M⊙1\,M_{\odot} core-accretion model and NKMN_{\mathrm{KM}} stars from the 0.7 M⊙0.7\,M_{\odot} model, matching the number of FG and KM stars in our survey. All planets around the drawn model stars were then evaluated against the survey sensitivity using the appropriate FG and KM detection completeness maps. Since wide-orbit giant planets are scarce in the models, we perturbed the mass and semi-major axis of each planet by a Gaussian of width 0.20.2 dex in log-space to improve sampling of the population beyond 10 au10\,\mathrm{au}. We restricted this yield analysis to companions below 10 MJup10\,M_{\mathrm{Jup}}, on the assumption that more massive substellar companions would likely already have been detected around our targets by previous surveys, and to focus the analysis on the core accretion dominated regime. Above this mass threshold, gravitational instability is expected to contribute significantly, complicating the interpretation in terms of a single formation channel.

Detection was treated as a Bernoulli trial: for every injected planet, we drew a uniform random number in [0,1][0,1] and classified the planet as detected if the draw fell below its associated detection probability pdetp_{\mathrm{det}} given by the completeness map. The total number of detections in each realization was then taken as the sum of Bernoulli successes across all injected planets. Repeating this procedure 10410^{4} times yields a Monte-Carlo distribution of the expected number of detections for the full JWST survey, under a given evolutionary model and normalization choice (right panel of Figure 13).

We show the results of this analysis in Figure 13 for the sonora-flame-skimmer-ceq chemical equilibrium model with C/O = 0.5 and [Fe/H] = 0.0, which represents one of the most optimistic cases in terms of survey sensitivity. Under these atmospheric predictions, the mean expected number of detections below 10 MJup10\,M_{\mathrm{Jup}} out of our 97 FGKM stars is 0.7, 1.2, and 2.8, using the model, SHINE, and RV normalizations, respectively. The spread in predicted yields across the three normalizations reflects the significant uncertainty in the underlying demographics of gas giant planets at wide separations, with the RV normalization consistently producing the most optimistic predictions by anchoring the population to the comparatively higher occurrence rates measured at smaller separations. In contrast, the most pessimistic case corresponds to the non-equilibrium, high-metallicity atmospheres (sonora-flame-skimmer-neq with C/O = 2.5 and [Fe/H] = 2.0), where the reduced survey completeness under these atmospheric assumptions leads to 0.3, 0.5, and 1.2 mean expected number of detections under the same three normalizations.

Using parametric models for planetary populations

As a complementary analysis, we also estimated the survey yield using an empirical parametric model for gas giant demographics. We adopted the parametric model of M. R. Meyer et al. [56], which represents the substellar companion population as the sum of planet and brown-dwarf components. Both components follow power-law distributions in the companion-to-star mass ratio, q=Mcomp/M⋆q = M_{\mathrm{comp}}/M_{\star}, and log-normal distributions in semimajor axis, aa. The model parameters were constrained using published companion-frequency measurements from radial velocity and direct imaging surveys. The brown dwarf component is truncated below 10 MJup10\,M_{\mathrm{Jup}} and therefore produces no companions within the mass range considered in this study, Mp≤10MJupM_{\mathrm{p}} \leq10 M_{\mathrm{Jup}}. The resulting yield is therefore entirely attributable to the planet component. In this model, the planet semimajor-axis distribution is truncated at 200 au to encompass the wide-separation planet-forming region observed in extended protoplanetary disks (e.g., J. Bae et al. [6])

The original model predicts a rapid decline in planet occurrence at wide separations because of its log-normal semimajor-axis distribution. We also consider an alternative model that allows for a more extended population of sub-Jupiters (E. Bogat et al. [10]). For planets with masses below 1 MJupM_{\mathrm{Jup}}, this model retains the original log-normal semimajor-axis distribution within 10 au but transitions to a distribution that is uniform in log⁡(a)\log(a) from 10 to 200 au, while leaving the mass-ratio distribution unchanged.

We applied both population models to the same sample of 97 single FGKM stars used in § 5.7.1, restricting the generated companions to Mp≤10MJupM_{\mathrm{p}} \leq10 M_{\mathrm{Jup}} and a≤200a \leq200 au. For each star, we constructed a 10×1010 \times10 grid that is uniformly spaced in log⁡(a)\log(a) and log⁡(q)\log(q). The semi-major axis spans 1–200 au, while the mass ratio spans the range corresponding to planet masses of 0.1–10 MJupM_{\mathrm{Jup}} for the given host-star mass. We calculated the expected planet occurrence in each grid cell by integrating the adopted mass ratio power-law and semimajor-axis distributions over the cell boundaries. The number of planets in each cell was then drawn from a Poisson distribution, where the expectation value is the integrated occurrence rate. Each generated planet is assigned a mass and semimajor axis by sampling uniformly within the corresponding cell. The subsequent detection procedure was identical to that described above: each simulated planet was considered detected through a Bernoulli trial with a success probability equal to its detection probability.

The left panel of Figure 14 shows a randomly selected 0.25% subset of the generated planets, with detected planets indicated by star symbols. By construction, the log-flat prescription produces an enhanced population of sub-Jupiters beyond 10 au and consequently more detections in this region. Averaged over the 97 single FGKM stars, the integrated occurrence rate of sub-Jupiters between 10 and 100 au is approximately 5.8% for the original model and 12.1% for the log-flat model.

Comparison of the original parametric demographic model and the alternative log-flat model

Figure 14. Comparison of the original parametric demographic model and the alternative log-flat model for the sonora-flame-skimmer-ceq atmosphere with C/O = 0.5 and [Fe/H] = 0.0. Left: A randomly selected 0.25% subset of the planets generated for the survey yield calculation, overlaid on the same average detection-completeness map shown in the left panel of Figure 13. Circles indicate generated planets, while star symbols mark those selected as detections through the Bernoulli trials. The original and log-flat populations are shown in orange and magenta, respectively. The log-flat prescription increases the abundance of sub-Jupiters beyond 10 au. Right: Normalized distributions of the expected survey yield, similar to the right panel of Figure 13, for the original and log-flat parametric demographic models.

For the same two atmospheric models discussed in § 5.7.1—the sonora-flame-skimmer-ceq model with C/O = 0.5 and [Fe/H] = 0.0, and the sonora-flame-skimmer-neq with C/O = 2.5 and [Fe/H] = 2.0—we compare the yields predicted by the two population prescriptions. For the C/O = 0.5 and [Fe/H] = 0.0 atmosphere, the original parametric model predicts a mean yield of 2.3 detectable planets among the 97 FGKM stars, whereas the log-flat model increases the mean yield to 4.7 planets. The right panel of Figure 14 shows the corresponding distributions of the expected yield across the Monte Carlo simulations. The log-flat distribution is shifted toward larger yields, reflecting the larger population of wide-orbit sub-Jupiters. For the high-metallicity atmospheres, the predicted mean yield increases from 0.8 planets under the original model to 1.1 planets under the log-flat model.

Limitations and prospects

A complete picture of the expected yields across all the BEX-ATMO and Sonora Flame Skimmer family models explored in this manuscript is presented in Table 6. The predictions are relatively stable across atmospheric model choices, with differences driven primarily by the underlying assumptions used to normalize the synthetic populations, ranging from <1<1 expected detection under the NGPPS core accretion normalization to ∼4\sim4 under the parametric log-flat model. The sonora-flame-skimmer-neq is overall the worst case due to their significantly reduced completeness, due to the NEQ models more aggressively suppressing flux than the CEQ models in F444W (see § 5.5.1 for reference), that result in a lowered expected numbers of detections.

ModelCO/FeHNGPPSSHINERVParametric (original)Parametric (log-flat)
bex-atmo2023-ceq0.55/0.00.61.22.62.14.4
bex-atmo2023-neq-s0.55/0.00.61.12.41.94.2
bex-atmo2023-neq-w0.55/0.00.61.12.41.84.1
sonora-flame-skimmer-ceq0.5/0.00.71.22.82.34.7
0.5/1.00.71.22.72.24.6
0.5/2.00.71.22.72.14.4
1.0/0.00.71.22.82.34.7
1.0/1.00.71.22.72.24.5
1.0/2.00.61.12.62.14.4
2.5/0.00.71.32.72.24.8
2.5/1.00.61.22.62.14.4
2.5/2.00.51.02.21.73.4
sonora-flame-skimmer-neq0.5/0.00.61.02.41.83.8
0.5/1.00.40.81.81.21.9
0.5/2.00.50.81.91.32.2
1.0/0.00.51.02.21.63.4
1.0/1.00.40.71.61.11.7
1.0/2.00.40.81.81.32.1
2.5/0.00.50.92.01.42.8
2.5/1.00.30.61.30.91.2
2.5/2.00.30.51.20.81.1

Table 6. Summary of the expected yields for Jupiter mass exoplanets for our programs combined, based on constrains from different synthetic population

The large disparity between normalization choices reflects a fundamental limitation of any yield prediction in this regime: the region of parameter space probed by our three programs, down to sub-Jupiter companions beyond ∼10\sim10 au, has not previously been constrained by any direct imaging surveys or radial velocity campaigns.

As a result, all yield predictions must rely either on theoretical models that lack empirical calibration in this regime, or on extrapolations from observed demographics of higher-mass or closer-in companions. The spread in predicted yields depending on the adopted population distribution and normalization reveals that current formation models and measured occurrence rates cannot yet be reconciled into a coherent picture at wide separations and low companion masses. This inconsistency is itself a valuable result, as it underscores the need for direct observational constraints in this regime to discriminate between competing predictions and improve our global understanding of wide-orbit giant planets.

The GO-4050, GO-5835, and SURVEY-6005 campaigns were specifically designed to address these demographic questions. The upcoming GO-10764 program will shortly validate or refute the initial candidate detections from these surveys. Subsequently, matching these empirical results with model predictions (in terms of numbers and properties) will provide us with a wealth of new information about the underlying mechanisms behind wide-orbit Jupiter and sub-Jupiter exoplanets, and will establish a critical benchmark for future demographic frameworks.

Conclusion

In this manuscript, we presented the results from the sensitivity limits analysis performed for three, recent, JWST programs GO-4050 and GO-5835 ([19] (A. Carter et al. 2023)), and SURVEY-6005 ([9]).

These three surveys have been carefully designed to survey nearby star forming regions and moving groups to provide a statistical sample of sub-Jupiter-mass exoplanets on wide orbits for demographic study.

We find the following key results:

  1. For JWST, the absolute magnitude sensitivity floor reached is governed by the instrument background noise rather than coronagraphic performance. GO-4050 and 5835 (mostly focused on M-dwarfs) exhibit poorer contrast while converging to identical magnitudes limits as SURVEY-6005 (mostly FGK stars). Moreover, SURVEY-6005’s superior contrast did not improve its absolute magnitude sensitivity at small separations, primarily because its FGK hosts are intrinsically brighter than the M dwarfs in GO-4050/5835.

  1. SURVEY-6005 exhibits a marginally higher limiting background compared to GO-4050/5835, as shorter readout ramps are required to prevent saturation for the brighter FGK stars observed in SURVEY-6005 compared to the fainter M stars observed in GO-4050/5835. This causes higher read noise compared to the longer DEEP8 ramps used in GO-4050/5835. These findings suggest that the necessary detector configuration for brighter targets limits the achievable sensitivity in the background-limited regime.

  1. For the GO-5835 imaging component, the absence of a coronagraph increased the overall throughput and signal-to-noise ratio achieved for this faint M-dwarf sample. This yielded improved performance at closer separation, and a background magnitude limit in F444W that is ∼1\sim1 magnitude deeper than the coronagraphic limits (∼22.2\sim22.2 at ≳2\gtrsim2 arcsec).

  1. Model selection has a strong impact on the minimum mass/TeffT_{\mathrm{eff}} achievable sensitivity limits. Nonetheless, using Sonora Flame-Skimmer models, we find that our programs are sensitive to sub-Saturn to Jupiter mass exoplanets with effective Temperature of ≲275 K\lesssim275\,K, already at ∼2\sim2 arcsec;

  1. We mapped the NIRCam/MIRI results from [11] onto our parameter space, finding consistent TeffT_{\mathrm{eff}} constraints: NIRCam demonstrates superior sensitivity relative to MIRI at distances ≳20\gtrsim20–30 pc. A comparison between the cloudy cases for the Sonora Flame Skimmer CEQ and NEQ models shows that while clouds may have a suppressive effect for NIRCam coronagraphic observations, increasing metallicity

might have a greater potential to limit F444W observations, depending on the atmospheric chemistry of the exoplanet.

  1. Comparing detection probability maps produced with the Sonora Flame Skimmer CEQ and NEQ models reveals that the detection probability for sub-Jupiter exoplanets is highly sensitive to the choice of equilibrium versus disequilibrium chemistry and model parameters, as well as the spectral type of the hosting star. For separations between ∼50\sim50 and 500 au, the model-to-model variation in F444W detection probability peaks at ∼50%\sim50\% for masses between ≲1\lesssim1 MJ_{\rm J} and ∼1\sim1 MSat_{\rm Sat}, and reaches ∼75%\sim75\% for effective temperatures of ∼300\sim300–400 K.

  1. Comparing detection probability maps generated for the K and M spectral type sample with similar maps for the AFG spectral type sample, we observe an increase in sensitivity towards the lowest masses for longer wavelengths with increased detection probabilities of ∼50%\sim50\% in favor of the K and M group for ≲1\lesssim1 MJ_{\rm J} at separation ≳20\gtrsim20 au, and ∼75%\sim75\% at ≳60\gtrsim60 au, a stark improvement compared to ground-based adaptive optics observations, thanks to the enhanced background limit from space with JWST compared to ground-based observations;

  1. PanCAKE simulations accurately reproduce NIRCam coronagraphic performance, with discrepancies between expected and empirical sensitivities remaining ≲10%\lesssim10\% at separations ≳30\gtrsim30 au, while at closer separations our programs perform ∼25%\sim25\% better compared to PanCAKE predictions.

  1. A comparison with the SHINE survey confirms that the improvements in mass sensitivity provided by JWST outpace ground-based observations, opening up a completely new parameter space to explore. In particular, in the region between few MJ_{\rm J} to ∼1\sim1 MSat_{\rm Sat}, JWST is able to reach improved detection probabilities by up to ∼50%\sim50\% compared to current ground observations at separations ∼50\sim50–500 au;

  1. Exoplanet yield predictions for the survey vary significantly depending on the underlying giant exoplanet population considered; this survey itself will be critical to refining our understanding of this population.

To conclude, we want to highlight the crucial role of large surveys (>100>100 stars) to build our understanding of the demographics of exoplanets. We demonstrate that the surveys presented here yield deep direct imaging sensitivity to sub-Jupiter exoplanets for the first time. Via ongoing follow-up of the candidate exoplanet detections revealed by this novel sensitivity, these surveys will provide the first statistically significant constraints on the frequency of sub-Jupiter exoplanets at semi-major axes of 30–200 AU.

Even though these are just predictions, and the only way to truly understanding the final mass/TeffT_{\rm eff} sensitivities is to follow-up candidate detections, study their atmospheric properties, and provide a direct constraint of which models are best at describing the reality of the observations, this work still provides the framework in which these detections must be interpreted. With enough targets to investigate multiple scenarios, we were able to deliver a detailed description on how variations in ages, distances, spectral-types, and membership (i.e. the environment in which exoplanet might form) would affect the predicted sensitivities and the nature of possible candidates, while the shared set-up minimize at the same time possible sources of bias.

Acknowledgments

BB and BJS acknowledge funding by the UK Science and Technology Facilities Council (STFC) grant nos. ST/V000594/1 and UKRI1196. This work is based on observations made with the NASA/ESA/CSA JWST. The data were obtained from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-03127 for JWST. These observations are associated with program nos. GO 4050 (PI: Carter), GO 5835 (PI: Carter), and SURVEY 6005 (PI: Biller). This research took advantage of the spaceKLIP ([35, 38] (A. L. Carter et al. 2023)) pipeline, and the pyKLIP tools ([83]), to support the initial data reduction. This research utilized the MADYS tools ([76]) to generate the sensitivity limits and detection probability maps on which this work is based on. This research made use of the Montreal Open Clusters and Associations (MOCA) database, operated at the Montréal Planétarium ([33]).

Author Contributions

Dr. G. M. Strampelli led the analysis of the sensitivity limits and detection probabilities on which this paper is based, as well as the reduction of the data and the assembling of the PSF library for the imaging dataset. Dr. A. Carter and Professor B. Biller led the three surveys GO-4050, 5835, and 6005 on which this work is based. Dr. C. Fontanive, Professor M. Meyer, and S. Han led the yield analysis. J. Mang led the development of the specific Sonora Flame Skimmer models used in this work. R. Ferrer-Chavez led the reduction of the data and the assembling of the PSF library for the coronagraphic dataset. Dr. V. Squicciarini developed and adapted the MADYS tools that provided the framework for this study. All the other members of the GO-4050, 5835, and 6005 collaboration equally contributed with invaluable feedback.

Facilities: JWST(NIRCam)

All the JWST data used in this paper can be found in MAST: 10.17909/3txt-jb91.

Software: spaceKLIP ([35, 38] (A. L. Carter et al. 2023)), PanCAKE ([16]), MADYS ([76]), MOCA ([33]) This research made use of Photutils, an Astropy package for detection and photometry of astronomical sources ([13]).

The target catalog

Simbad ID2Mass IDDistanceJKF444WContrastMembershipGOSpTE(B-V)
BD+22 440921310171+232007424.218±0.01024.218 \pm0.0107.0756.3826.202.14e-06ABDMG6005K8V0.001±0.0150.001 \pm0.015
CD-26 164304213348-254309154.566±0.05654.566 \pm0.0567.1936.8736.872.79e-06ABDMG6005F9V0.0±0.0140.0 \pm0.014
CD-40 170105023042-395912955.217±0.03155.217 \pm0.0318.7278.058.046.97e-06ABDMG6005K4V0.0±0.0140.0 \pm0.014
HD 4527006223097-601307223.873±0.00823.873 \pm0.0085.4335.0455.053.62e-06ABDMG6005G1V0.0±0.0140.0 \pm0.014
HD 2595304064153+014102057.082±0.08657.082 \pm0.0866.8926.5826.581.15e-04ABDMG6005F5V0.001±0.0140.001 \pm0.014
HD 6151807370278-522157061.354±0.05961.354 \pm0.0596.9686.696.662.48e-06ABDMG6005F5V0.001±0.0140.001 \pm0.014
BD+41 474923193955+421509851.290±0.04151.290 \pm0.0417.6057.2247.173.84e-06ABDMG6005G0V0.004±0.0140.004 \pm0.014
HD 22422823561067-390308621.954±0.01121.954 \pm0.0116.515.9075.881.77e-06ABDMG6005K2V0.0±0.0150.0 \pm0.015
HD 21734323001929-260913531.777±0.02231.777 \pm0.0226.2915.9435.901.76e-06ABDMG6005G3V0.001±0.0150.001 \pm0.015
BD+05 457620395460+062011836.563±0.12236.563 \pm0.1227.9227.1337.023.30e-06ABDMG6005K7Ve0.001±0.0150.001 \pm0.015
V* HS Psc01372322+265711937.755±0.02637.755 \pm0.0268.4297.6427.563.50e-02ABDMG6005K5V0.001±0.0140.001 \pm0.014
HD 20191921130526-172912638.431±0.02438.431 \pm0.0248.3487.5837.534.46e-06ABDMG6005K6V0.001±0.0150.001 \pm0.015
HD 3565005243016-385810617.461±0.00517.461 \pm0.0056.7025.9215.782.36e-06ABDMG6005K6V0.0±0.0140.0 \pm0.014
HD 15255516540815-042024845.196±0.04745.196 \pm0.0476.76.3636.353.12e-04ABDMG6005F9:V0.001±0.0160.001 \pm0.016
EROS-MP J0032-440500325584-440505834.551±0.50434.551 \pm0.50414.77613.26912.409.87e-05ABDMG5835L0γ\gamma0.001±0.0150.001 \pm0.015
BD+20 179007234358+202458827.712±0.01427.712 \pm0.0147.6436.8796.842.96e-06ABDMG6005K3(V)0.0±0.0140.0 \pm0.014
BD+05 37802412589+055918144.296±0.04244.296 \pm0.0427.9047.0696.943.50e-06AUSC5835M00.001±0.0140.001 \pm0.014
BD-03 557923093711-022555152.018±0.05252.018 \pm0.0528.5677.8187.745.54e-06AUSC6005K4Ve0.001±0.0150.001 \pm0.015
2MASS J06085283-275358306085283-275358344.209±0.30744.209 \pm0.30713.59512.37111.557.20e-05AUSC5835L0δ\delta0.0±0.0140.0 \pm0.014
RX J0019.7+195100194303+195111757.471±0.07257.471 \pm0.07210.7189.8649.341.63e-05AUSC5835M40.003±0.0140.003 \pm0.014
BD-13 642423323085-121551327.452±0.01627.452 \pm0.0167.456.5696.373.84e-06BPMG5835M00.0±0.0140.0 \pm0.014
2MASS J19243494-344329219243494-344329251.547±0.08251.547 \pm0.0829.678.7858.306.62e-06BPMG5835M4.60.003±0.0160.003 \pm0.016
PM J19560800-320?19560800-320?27.630±0.04727.630 \pm0.0478.718.0677.599.78e-06BPMG5835M40.0±0.0140.0 \pm0.014
RX J0506.2+043905061292+043927251.452±0.06151.452 \pm0.0619.098.5848.214.44e-06BPMG5835M3.50.006±0.0140.006 \pm0.014
1RXS J184260.6-55542618426064-55542641.654±0.05441.654 \pm0.0549.4888.5849.217.28e-06BPMG5835M3.50.006±0.0160.006 \pm0.016
UCAC4 171-191-08321000535-191957332.365±0.05732.365 \pm0.0578.6777.9566.931.86e-06BPMG5835M4.60.001±0.0150.001 \pm0.015
BPS CS 22898-006500501052+083734161.308±0.19661.308 \pm0.1969.7458.8978.342.87e-06BPMG5835M20.002±0.0140.002 \pm0.014
PM J00502+083721000461-284303233.619±0.70633.619 \pm0.7068.4287.5127.087.43e-06BPMG5835M50.002±0.0170.002 \pm0.017
UCAC4 354-?01112542+156?21.933±0.85921.933 \pm0.8599.0828.2087.717.43e-05BPMG5835M8.50.0±0.0140.0 \pm0.014
2MASS J193526-28464321374019+013713735.913±0.16135.913 \pm0.1618.8027.8847.462.80e-03BPMG5835M5.50.001±0.0150.001 \pm0.015
LP 467-1622447594-361501520.867±0.01120.867 \pm0.0117.7866.9826.487.29e-06BPMG5835M40.0±0.0150.0 \pm0.015
RX J2137.6+013700172333-664512436.823±0.02336.823 \pm0.0238.5637.7037.393.89e-06BPMG5835M30.001±0.0150.001 \pm0.015
Simbad ID2Mass IDDistanceJKF444WContrastMembershipGOSpTE(B-V)
SIPS J2335-340123355015-340147737.376 ±\pm 0.06311.63910.75810.172.87e-05BPMG5835M60.001 ±\pm 0.015
PM J22088+114422085034+114413137.388 ±\pm 0.0389.9019.0378.548.52e-06BPMG5835M4.30.001 ±\pm 0.015
SIPS J1809-761318090694-761323927.277 ±\pm 0.0169.8178.9898.437.52e-06BPMG5835M50.002 ±\pm 0.015
V* AO Men06182824-720241639.115 ±\pm 0.0197.536.8146.732.56e-06BPMG5835K3.50.001 ±\pm 0.015
HD 20300065008-230627139.741 ±\pm 0.0545.4515.245.233.73e-06BPMG6005F30.001 ±\pm 0.014
BD+30 39702272924+305824640.945 ±\pm 0.0257.877.087.043.70e-06BPMG5835K80.001 ±\pm 0.014
CD-31 1604118504448-314747249.457 ±\pm 0.0338.3147.4627.364.79e-06BPMG6005K80.002 ±\pm 0.016
1RXS J203336.9-25565420333759-255652143.662 ±\pm 0.0779.7128.8778.389.09e-06BPMG5835M4.80.002 ±\pm 0.016
HD 20704321475527-525550233.243 ±\pm 0.0196.4156.056.027.99e-06BPMG5835G5V0.002 ±\pm 0.016
PM J21214-665521212873-665506331.563 ±\pm 0.5747.887.0096.924.74e-06BPMG5835M00.002 ±\pm 0.015
2MASS J22334687-295010122334687-295010152.124 ±\pm 0.27012.94111.97111.397.30e-05BPMG5835M60.001 ±\pm 0.015
UPM J1354-712113545390-712147622.805 ±\pm 0.0108.5497.6727.365.06e-06BPMG5835M2.50.001 ±\pm 0.015
SIPS J2000-752320004841-752307029.403 ±\pm 0.09412.73411.51110.836.00e-05BPMG5835M90.002 ±\pm 0.015
2MASSW J0335020+23423503350208+234235650.702 ±\pm 0.23112.2511.26110.684.50e-05BPMG5835M8.50.001 ±\pm 0.014
2MASS J19260075-533126919260075-533126947.760 ±\pm 0.0649.5958.6848.227.37e-06BPMG5835M40.006 ±\pm 0.016
Smethells 2018465255-621036650.700 ±\pm 0.0488.7467.8547.689.04e-06BPMG5835M1.50.005 ±\pm 0.016
CD-57 105405004714-571525526.873 ±\pm 0.0097.0956.2446.072.35e-06BPMG5835M00.0 ±\pm 0.014
HD 17316718480637-621347050.571 ±\pm 0.0566.4166.1366.112.56e-06BPMG6005F50.005 ±\pm 0.016
2MASS J19300396-293932219300396-293932260.253 ±\pm 0.08410.1179.2538.781.21e-05BPMG5835M40.003 ±\pm 0.017
Barta 161 1201351393-071251737.281 ±\pm 0.0648.9648.0787.693.99e-06BPMG5835M40.0 ±\pm 0.014
2MASS J02241739+203151302241739+203151370.753 ±\pm 0.36312.4811.61911.004.67e-05BPMG5835M60.005 ±\pm 0.014
GR* 900281434-322755635.008 ±\pm 0.02910.1219.2788.738.90e-06BPMG5835M50.0 ±\pm 0.014
GJ 2006 A00275023-323306035.029 ±\pm 0.0488.8828.0127.495.24e-06BPMG5835M3.50.0 ±\pm 0.014
RX J0534.0-022105335981-022132534.346 ±\pm 0.0328.5647.6957.333.98e-06BPMG5835M30
V* V1005 Ori04593483+014700724.396 ±\pm 0.0087.1176.2616.082.77e-06BPMG5835M30.001 ±\pm 0.014
HD 35862053?45.818 ±\pm 0.2077.8497.0396.999.53e-03BPMG6005K7V0.001 ±\pm 0.015
RX J0613.2-2742061333?33.772 ±\pm 0.2498.0027.4166.801.85e-05BPMG5835M3.80.0 ±\pm 0.014
RX J0629.7-3233062944?59.915 ±\pm 0.01711.72510.83110.291.23e-04BPMG5835M5.30.003 ±\pm 0.017
2MASS J18473151-280855818473151-280855858.822 ±\pm 0.0427.5326.9446.776.47e-03CAr6005K00.003 ±\pm 0.015
HD 42270191?58.137 ±\pm 0.0317.5766.9816.863.92e-06CAr6005K10.002 ±\pm 0.014
HD 49427064346?55.389 ±\pm 0.0447.6937.2787.213.31e-06CAr6005G60.003 ±\pm 0.014
HD 4985507003046-794145963.435 ±\pm 0.0548.2657.6527.661.89e-05CAr6005K20.003 ±\pm 0.015
S1 392143?30.972 ±\pm 0.0118.2277.3797.182.77e-04CARN6005K70.001 ±\pm 0.014
2MASS J0102-630602010?42.467 ±\pm 0.36713.44812.94412.211.51e-04CARN6005L30.001 ±\pm 0.014
HD 6285007423606-591750732.856 ±\pm 0.0176.0725.7015.711.58e-06CARN6005G2.50.001 ±\pm 0.014
HD 6360807461619-494383433.764 ±\pm 0.0216.8626.4116.273.19e-06CARN6005K0V0.001 ±\pm 0.014
1RXS J040716.4-691823040716?56.962 ±\pm 0.03610.4578.1888.031.19e-04COL5835F40.001 ±\pm 0.014
V* AS Col052?47.340 ±\pm 0.0346.4166.1446.101.85e-06COL6005F60.0 ±\pm 0.014
Simbad ID2Mass IDDistanceJKF444WContrastMembershipGOSpTE(B-V)
HD 4107106004130-445350054.113±0.03454.113 \pm0.0347.737.3217.531.67e-03COL6005G80.0±0.0140.0 \pm0.014
HD 3839705433580-395524653.594±0.04953.594 \pm0.0497.0566.7566.742.79e-06COL6005G00.0±0.0140.0 \pm0.014
HD 4021605554314-380616252.767±0.04452.767 \pm0.0446.4946.2066.162.86e-03COL6005F70.0±0.0140.0 \pm0.014
2MASS J03550477-103241503550477-103241552.264±0.46652.264 \pm0.46613.0811.97511.321.48e-03COL5835M8.30.0±0.0140.0 \pm0.014
PM J05501+051105500858+051153664.060±0.07764.060 \pm0.0779.3698.4718.258.05e-06COL5835M20.001±0.0140.001 \pm0.014
UCAC4 345-00684205082729-210144448.304±0.13948.304 \pm0.1399.7168.8338.348.99e-06COL5835M50.0±0.0140.0 \pm0.014
V* V1358 Ori06190805-032620351.672±0.06051.672 \pm0.0606.8486.5526.492.23e-06COL6005G0V0.0±0.0140.0 \pm0.014
CD-48 479709331427-484833146.278±0.02446.278 \pm0.0248.9458.1018.027.69e-06FCCN6005K7V0.001±0.0150.001 \pm0.015
e Vir13164653+092526917.588±0.03817.588 \pm0.0384.3924.0333.752.28e-06FIELD6005G0IV0.0±0.0150.0 \pm0.015
V* LV Com12075091+185655841.963±0.03741.963 \pm0.0377.5447.02810.006.15e-05FIELD6005K2V0.001±0.0140.001 \pm0.014
2MASS J02495639-055735202495639-055735266.251±0.25066.251 \pm0.25011.96311.05510.433.49e-03FIELD5835M60.001±0.0140.001 \pm0.014
V* AF Psc23314492-024439534.927±0.04334.927 \pm0.0439.5078.6678.136.67e-06FIELD5835M50.001±0.0140.001 \pm0.014
HD 22123923301372+314221025.086±0.01425.086 \pm0.0146.666.1456.071.70e-06FIELD6005K2.5V0.001±0.0140.001 \pm0.014
PM J13219-105213215631-105209850.396±0.78350.396 \pm0.7839.5248.6188.261.16e-05FIELD5835M3.5Ve0.001±0.0160.001 \pm0.016
MCC 12410141918+210429723.368±0.10323.368 \pm0.1037.0746.2616.077.10e-06FIELD5835M1V0.0±0.0140.0 \pm0.014
1RXS J112755.0-66255811275535-6626046109.246±0.141109.246 \pm0.1419.0068.3588.297.32e-05MEL46005K1V(e)0.016±0.0170.016 \pm0.017
TWA 4612354615-411553152.931±0.17252.931 \pm0.17210.0689.2328.771.06e-05OCTN4050M30.003±0.0160.003 \pm0.016
CPD-63 128610065573-6352086136.533±1.272136.533 \pm1.2729.2628.588.461.25e-05SCOCEN6005K0V(e)0.014±0.0170.014 \pm0.017
CD-53 54402414683-525952343.598±0.08743.598 \pm0.0877.5826.7636.682.74e-06THA6005K60.001±0.0140.001 \pm0.014
CD-78 2400422033-774739749.792±0.02449.792 \pm0.0248.227.5297.494.00e-06THA6005K3.50.002±0.0150.002 \pm0.015
HD 2463603481148-744138857.112±0.04957.112 \pm0.0496.3676.1366.122.33e-06THA6005F30.002±0.0150.002 \pm0.015
HD 3219504480518-804645262.474±0.06862.474 \pm0.0687.176.8686.872.79e-06THA6005F70.002±0.0150.002 \pm0.015
Smethells 8621443012-605838946.414±0.03046.414 \pm0.0308.7517.8747.724.81e-06THA6005K80.003±0.0150.003 \pm0.015
HD 20757521520973-620308546.761±0.03246.761 \pm0.0326.3586.0276.032.93e-06THA6005F60.003±0.0150.003 \pm0.015
UCAC3 13-3121723585674-833942355.635±0.03855.635 \pm0.0389.1218.2468.131.20e−061.20e-06THA6005K5.80.003±0.0150.003 \pm0.015
CD-86 14723274943-861318763.366±0.04663.366 \pm0.0467.957.458.134.19e-06THA6005G80.003±0.0150.003 \pm0.015
HD 2094721200496-742003046.709±0.03446.709 \pm0.0347.3866.9086.943.42e-06THA6005G60.002±0.0150.002 \pm0.015
HD 98700077414-741141157.064±0.02457.064 \pm0.0247.4066.9626.862.77e-06THA6005G70.002±0.0150.002 \pm0.015
HD 1726002461462+053533357.043±0.07857.043 \pm0.0786.8596.5176.423.64e-03THA6005F70.001±0.0140.001 \pm0.014
2MASS J04231498-153324504231498-153324544.420±0.04244.420 \pm0.0427.3376.53311.273.87e-03THA5835K7.50.001±0.0150.001 \pm0.015
HD 351?04351?51.907±0.08551.907 \pm0.0857.2036.8756.403.02e-06THA6005K30.0±0.0140.0 \pm0.014
HD 23402040083?38.830±0.03238.830 \pm0.0326.4646.1176.102.22e-06THA6005G00.001±0.0150.001 \pm0.015
HD 10500055255-414510946.131±0.22046.131 \pm0.22010.6757.7639.261.69e-05THA5835M5.80.001±0.0140.001 \pm0.014
SCR J0336-261903361478-261957844.183±0.02444.183 \pm0.02411.96410.86410.294.10e-06THA5811M7.80.001±0.0150.001 \pm0.015
DENIS J004135.3-56211200413533-56211236.789±0.95836.789 \pm0.95811.96110.86110.296.47e-05THA5835M7.80.001±0.0150.001 \pm0.015
HD 8588012321?45.268±0.03245.268 \pm0.0327.2416.8476.753.21e-06THA6005G60.001±0.0150.001 \pm0.015
2MASS J021?020?45.279±0.03545.279 \pm0.0358.4387.5639.651.22e-06THA6005M3.80.001±0.0140.001 \pm0.014
CD-35 1167031983?45.407±0.03445.407 \pm0.03410.5767.7237.614.86e-06THA6005K30.0±0.0140.0 \pm0.014
BD-15 70504021648-151229754.700±0.03954.700 \pm0.0398.2287.5737.544.25e-06THA6005K3.40.0±0.0140.0 \pm0.014
Simbad ID2Mass IDDistanceJKF444WContrastMembershipGOSpTE(B-V)
V* DK Cet01574896-215405241.313 ± 0.0426.8566.4726.492.10e-06THA6005G40.0 ± 0.014
HD 1026901390761-562545946.925 ± 0.0366.3355.9895.971.77e-06THA6005F50.001 ± 0.015
HD 1086301460105-272055644.892 ± 0.0625.7235.4565.412.09e-06THA6005F0IV0.0 ± 0.014
CD-46 106403304909-455557343.895 ± 0.0207.7727.1037.072.80e-06THA6005K30.0 ± 0.014
PM J05334-511705332558-511713153.880 ± 0.0348.9868.1588.018.81e-06THA6005K70.0 ± 0.014
HD 2961504384393-270201854.349 ± 0.0447.2736.8666.873.47e-06THA6005G30.0 ± 0.014
V* V1252 Cen12350424-413638557.483 ± 0.0859.1228.1868.297.55e-06TWA4050M30.003 ± 0.016
HD 29893610131476-523054053.085 ± 0.0317.877.1947.165.74e-06TWA6005K3Ve0.003 ± 0.016
TWA 3311393382-304000248.903 ± 0.0619.9859.1188.458.72e-06TWA4050M4.750.001 ± 0.016
SCR J1237-402112371238-402148063.601 ± 0.1019.4738.5158.196.23e-06TWA4050M2.5Ve0.003 ± 0.016
CD-36 7429A11482422-372849176.489 ± 0.0818.6847.8487.712.11e-03TWA4050K7IVe0.003 ± 0.016
HD 10957312360103-395210270.773 ± 0.2365.7845.7695.827.51e-04TWA4050A00.004 ± 0.016
10 Crt11084400-280450454.606 ± 0.2245.3135.2355.242.58e-06TWA4050A20.001 ± 0.015
TWA 2012313807-455859380.207 ± 0.1559.3318.4128.126.03e-06TWA4050M3IVe0.006 ± 0.016
V* V1217 Cen11210549-384516365.404 ± 0.0828.9998.0537.866.20e-06TWA4050M1Ve0.002 ± 0.016
TWA 4511592786-451019272.258 ± 0.4909.9349.0658.649.44e-06TWA4050M4.50.004 ± 0.016
CD-34 7390A11211723-344645459.847 ± 0.0598.4317.4917.404.78e-03TWA4050M1Ve0.001 ± 0.016
TWA 3410284580-283037461.176 ± 0.13410.95310.0269.031.92e-05TWA4050M6γ0.001 ± 0.014
TWA 3011321831-301951847.419 ± 0.0999.6418.7658.155.82e-06TWA4050M4.5Ve0.001 ± 0.015
TWA 2611395113-315921446.722 ± 0.51212.68611.50310.786.56e-05TWA4050M9γ0.001 ± 0.015
2MASS J09462782-445740809462782-445740846.646 ± 0.06810.8769.9929.421.85e-05TWA5835M5.50.001 ± 0.015
CD-26 862311324124-265155945.959 ± 0.0608.3377.437.188.08e-06TWA4050M2Ve0.001 ± 0.015
UCAC3 133-13934110585054-234620643.796 ± 0.05810.3029.438.871.22e-05TWA4050M6γe0.001 ± 0.015
V* V1249 Cen12153072-394842653.604 ± 0.0748.1667.3067.162.89e-06TWA4050K9IV-Ve0.002 ± 0.016
TWA 3512002750-340537170.765 ± 0.3749.6068.7238.281.17e-05TWA4050M40.002 ± 0.016
TWA 4412175920-373443377.504 ± 0.22911.57210.71610.193.96e-05TWA4050M5e0.002 ± 0.016
V* E? Ant.10442301-442907734.098 ± 0.03214.18213.89912.252.76e-04TWA4050M3.250.006 ± 0.017
TWA 2912023799-383280262.576 ± 0.11210.6689.8479.361.80e-05TWA4050M4.80.002 ± 0.016
TWA 3612072839-324700256.822 ± 0.2888.6187.7517.464.21e-05TWA4050M2.5Ve0.002 ± 0.016
TWA 2307300559-371347532.680 ± 0.0096.1095.5895.161.96e-06UMA6005G5V0.0 ± 0.014
HD 5974707261067+673942525.811 ± 0.0136.7846.2466.162.88e-06UMA6005K0V0.0 ± 0.014
HD 5616812122084+529086815.249 ± 0.0036.8756.0595.861.95e-06UMA6005K7V0.0 ± 0.014
HD 23809008550438+361146425.389 ± 0.0146.2585.8835.822.69e-06UMA6005G90.0 ± 0.014
HD 7621809262296+416421922.075 ± 0.0146.5155.1775.197.63e-06UMA6005G0V0.0 ± 0.014
HD 7599004152879+061112822.071 ± 0.0125.2014.9034.872.30e-06UMA6005G0V0.0 ± 0.014
HD 2692304152578+061158822.071 ± 0.0125.6955.2715.402.73e-06UMA6005G8V0.0 ± 0.014
HD 2691304140417+053231715.777 ± 0.0075.3174.8224.832.18e-06UMA6005K0V0.001 ± 0.014
V* V1368 Ori12414417+053238815.722 ± 0.0116.5696.0035.981.98e-06UMA6005K3V0.001 ± 0.014
HD 11046118062370-360111317.115 ± 0.0124.8354.4694.485.96e-06UMA6005G1V0.001 ± 0.015
Simbad ID2Mass IDDistanceJKF444WContrastMembershipGOSpTE(B-V)
HD 21781323030499+205506824.295±0.01424.295 \pm0.0145.4925.1485.143.93e-06UMA6005G1V0.001±0.0140.001 \pm0.014
HD 11504313133699+564229825.161±0.01025.161 \pm0.0105.6755.3345.325.08e-04UMA6005G1V0.001±0.0140.001 \pm0.014
HD 18496019341977+511412125.431±0.01825.431 \pm0.0184.74.4944.501.92e-06UMA6005F7V0.001±0.0150.001 \pm0.015
18 Boo14191627+130015526.106±0.05226.106 \pm0.0524.7394.3944.392.33e-06UMA6005F5IV0.001±0.0150.001 \pm0.015
HD 16738918130721+412831534.674±0.01734.674 \pm0.0176.2245.9185.891.45e-06UMA6005F8(V)0.003±0.0150.003 \pm0.015
HD 11145612483942+601911526.151±0.19926.151 \pm0.1995.1784.5554.501.44e-05UMA6005F5V0.001±0.0140.001 \pm0.014
HD 7593508534994+265447737.893±0.04237.893 \pm0.0427.1026.6536.692.71e-06UMA6005G8V0.001±0.0140.001 \pm0.014
V* AN CrB15353016+361234730.576±0.01330.576 \pm0.0137.0116.4969.393.82e-05UMA6005K0V0.002±0.0150.002 \pm0.015
V* DL Dra14420323+611542941.148±0.04341.148 \pm0.0435.4775.2465.211.72e-03UMA6005F2V0.004±0.0140.004 \pm0.014
BD-03 200207400289-033613014.086±0.00514.086 \pm0.0056.3775.5685.471.98e-06UMA6005K7V0.0±0.0140.0 \pm0.014
HD 2932904460056+763639932.579±0.02032.579 \pm0.0205.5355.2845.271.91e-06VUL6005F7V0.001±0.0140.001 \pm0.014

Table A1. Description of the 165 sources from Table 3. Each source is listed with its known Simbad ID, 2Mass ID, distance (in parsecs), J, K, and F444W magnitudes, contrast at 2″ in F444W, membership, GO number, spectral type, and extinction. All the values, with the exception of the contrast, membership, and Program number are obtained from MOCAdb. Ages are assumed based on their respective membership and reported in Table 3.

Accounting for Binarity in the Imaging Data

As shown by Table 2, ∼28%\sim28\% of our targets host a visually resolved binary companion. Binaries require a more curated approach, since the presence of a bright close-in visually resolved companion will heavily bias any attempt at correctly recentering and aligning the primary to the center of the frame. Additionally, they can also heavily bias the residuals and the final contrast curves, if light from the companion or disk is not carefully removed.

Thus, we must establish a new approach to handle them carefully. Even though binaries will be the focus of other papers (e.g. (Bruinsma et al. in prep.)), in this section we introduce a new approach implemented in spaceKLIP to recenter the three visually resolved binaries in the imaging dataset (2MASS-J04231498-1533245, 2MASS-J02495639-0557352, and 2MASS-J03550477-1032415).

For this purpose, we developed a new MCMC tool for the pipeline (taking advantage of emcee11, a Python implementation of Goodman & Weare’s Affine Invariant Markov chain Monte Carlo - MCMC - Ensemble sampler; D. Foreman-Mackey et al. [31]) to fit for two different PSFs simultaneously. The goal is to perform a double PSF-fitting to find the best match between the image coordinates and brightness of the primary, and the separation, position angle and contrast of the companion.

To do so, the tool allows the user to choose between two distinct approaches, using either STPSF (M. D. Perrin et al. [67, 66])12, or the Photutils EPSF module13. STPSF is a Python package that computes simulated point spread functions (PSFs) for NASA’s James Webb Space Telescope (JWST) and Nancy Grace Roman Space Telescope, taking into account detector pixel scales, rotations, filter profiles, and point source spectra. The EPSF module from Photutils instead, derives an empirical PSF, or effective PSF (ePSF), by using a set of real sources from the surveys to model what fraction of a star’s light will land in a particular pixel (J. Anderson [4]; J. Anderson & I. R. King [5]). Once the best fit is obtained, the retrieved image coordinates of the primary can be used to evaluate the shift needed to recenter the couple with the primary at the center of the frame.

Before we can estimate contrast limits for these targets, we need to mitigate the loss of sensitivity caused by the presence of the visual binary companion in the field of view (as shown by the left panel of Figure 15).

Calibrated contrast curves for isolated sources and binaries, and comparison of binary contrast curves before and after companion removal

Figure 15. Left: Calibrated contrast curve comparison between isolated sources (orange) and binaries (magenta) in the GO-4050, GO-5835, and SURVEY-6005 surveys combined. Right: Comparison between median calibrated contrast curves of the three binaries in the GO-5835 imaging dataset: 2MASS J04231498-1533245, 2MASS J02495639-0557352, and 2MASS J03550477-1032415, before (magenta continuous line) and after removing the visual companion using models generated with STPSF (magenta dashed lines), and Photutils EPSF module (magenta dashed-dotted lines). For comparison is also shown the median calibrated contrast curves obtained from all the other targets in the NIRCam imaging sample that are not visual binaries (orange continuous line). In both plots, the lines marks the median of the distribution, while the shaded area cover the 16th/84th percentile.

To do so, we took advantage of the extract_companions functionality of spaceKLIP Analysis tool, where we extracted the bright companion from the residual of each target binary, producing a new residual where the visual companion was greatly attenuated. Similar to what we describe in § 3.1 to perform double PSF fitting to recenter the three visual binaries, we model the PSF of the bright close-in companion using both the STPSF and EPSF.

In the right panel of Figure 15 we show the outcome of this process: the calibrated contrast curves of the three visual binaries in the GO-5835 imaging dataset pre- (continuous lines) and post-companion removal using two different methods to generate the PSF for the companion: STPSF (dashed lines) and Photuils EPSF module (dashed-dotted lines). For comparison, in the same figure we also show in orange the median calibrated contrast curves obtained from the other targets in the GO-5835 imaging sample. As expected, when the bright companion is removed, we see a net improvement in the contrast, more in line with the median contrast curve for isolated targets. Overall, once we remove the binaries, we reach the same ≲10−4\lesssim10^{-4} contrast level at separations ≳1−1.5\gtrsim1-1.5 arcsec of the GO-5835 isolated targets in the imaging dataset.

Group-based Detection Probability maps

Here we present in Figure 16 and Figure 17 detection probability maps for the most best-sampled groups in our programs (TWA, BPMG, COL, THA, ABDMG, and UMA). All maps are derived from the F444W sensitivity limits shown in Figure 3. Following results from § 5.6.1, we ordered these maps as a function of the age, from the younger group to the older. For each group, we report the most common spectral types (SpT) within that sample, close to the group name. Table A2 instead provide a more quantitative representation, showing the expected mass detection probability in the region 0.3–1 MJup_{\mathrm{Jup}} and 50–500 au, and the TeffT_{\mathrm{eff}} detection probability in the region 300–1000 K and 30–200 au, for each of the best-sampled groups.

Mass detection probability maps for two Sonora Flame Skimmer models

Figure 16. Mass detection probability maps comparison for two Sonora Flame Skimmer models, ordered by age (younger on the left). The most common spectral types (SpT) in each sample are labeled for each group. All maps are derived from the F444W sensitivity limits shown in Figure 3. The yellow and blue horizontal dashed lines show the Jupiter and Saturn reference masses.

Teff detection probability maps for two Sonora Flame Skimmer models

Figure 17. TeffT_{\mathrm{eff}} detection probability maps comparison for two Sonora Flame Skimmer models, ordered by age (younger on the left). The most common spectral types (SpT) in each sample are labeled for each group. All maps are derived from the F444W sensitivity limits shown in Figure 3. The black horizontal dashed line shows the Epsilon Ind AB reference temperature [54].

ModelCO/FeHMass TWAMass BPMGMass COLMass THAMass ABDMGMass UMATeffT_{\mathrm{eff}} TWATeffT_{\mathrm{eff}} BPMGTeffT_{\mathrm{eff}} COLTeffT_{\mathrm{eff}} THATeffT_{\mathrm{eff}} ABDMGTeffT_{\mathrm{eff}} UMA
bex-atmo2023-ceq0.55/0.093.6%83.2%77.5%79.4%49.4%11.2%93.6%95.9%93.4%94.6%96.6%96.0%
bex-atmo2023-neq-s0.55/0.093.6%83.2%78.1%79.4%49.5%11.4%92.6%95.0%92.3%92.4%95.8%95.1%
bex-atmo2023-neq-w0.55/0.093.7%83.2%78.0%79.8%49.4%11.3%92.7%94.6%91.9%92.5%94.8%94.1%
sonora-flame-skimmer-ceq0.5/0.095.2%84.3%81.2%81.3%47.3%6.8%94.3%96.5%95.1%95.5%97.0%95.3%
0.5/1.094.8%84.4%80.0%80.2%49.4%7.5%93.3%96.0%93.8%94.4%96.8%95.8%
0.5/2.092.9%83.4%79.4%80.6%48.9%8.3%81.1%89.8%82.2%85.1%92.3%92.0%
1.0/0.095.5%84.5%82.7%81.4%48.2%7.3%94.8%96.6%95.6%95.5%97.2%96.5%
1.0/1.094.7%84.1%78.9%80.2%49.9%8.1%93.4%96.0%93.6%94.7%96.8%94.6%
1.0/2.092.7%82.6%74.7%78.8%51.9%9.6%80.9%89.7%82.1%85.4%92.1%91.8%
2.5/0.094.6%84.3%80.9%81.8%49.5%8.5%94.3%96.6%94.7%95.6%97.1%96.7%
2.5/1.093.7%84.0%77.1%79.9%49.9%8.5%93.0%95.8%92.7%94.4%96.6%95.3%
2.5/2.089.9%79.4%64.5%66.1%29.6%3.0%79.6%89.3%82.1%84.5%92.4%91.0%
sonora-flame-skimmer-neq0.5/0.092.5%80.9%64.5%67.9%43.2%4.2%90.2%94.7%90.1%91.8%96.2%95.0%
0.5/1.075.2%50.5%20.0%22.6%6.9%0.0%79.6%87.1%79.0%82.2%90.2%89.5%
0.5/2.081.5%56.6%27.2%28.5%1.5%0.0%62.8%73.4%62.8%67.1%79.2%79.0%
1.0/0.090.0%78.6%50.0%53.6%36.6%4.9%89.6%94.2%88.4%90.3%94.7%94.0%
1.0/1.069.4%45.9%15.3%17.8%4.7%0.0%78.1%85.8%78.0%81.3%88.4%88.5%
1.0/2.076.2%49.8%21.8%22.4%2.7%0.0%62.5%74.5%64.5%68.0%79.0%77.6%
2.5/0.085.1%76.1%44.6%50.5%25.6%1.3%87.0%92.9%85.9%88.0%93.4%91.4%
2.5/1.051.5%28.5%3.4%7.7%0.2%0.0%72.8%80.8%72.1%75.1%84.7%80.0%
2.5/2.046.4%20.9%0.4%1.3%0.1%0.0%56.3%67.6%55.4%59.2%70.8%67.1%

Table A2. Mean detection probability (%) for mass (0.3–1.0 MJM_{\mathrm{J}}, 50–500 AU) and TeffT_{\mathrm{eff}} (300–1000 K, 30–200 AU).

References

  1. [1]Allard, F., Hauschildt, P. H., Alexander, D. R., Tamanai, A., & Schweitzer, A. 2001, ApJ, 556, 357, doi: 10.1086/321547DOI
  2. [2]Allard, F., Homeier, D., & Freytag, B. 2012, Philosophical Transactions of the Royal Society of London Series A, 370, 2765, doi: 10.1098/rsta.2011.0269DOI
  3. [3]Amara, A., & Quanz, S. P. 2012, MNRAS, 427, 948, doi: 10.1111/j.1365-2966.2012.21918.xDOI
  4. [4]Anderson, J. 2016, Instrument Science Report WFC3 2016-12, 42 pages
  5. [5]Anderson, J., & King, I. R. 2000, PASP, 112, 1360, doi: 10.1086/316632DOI
  6. [6]Bae, J., Teague, R., Andrews, S. M., et al. 2022, ApJL, 934, L20, doi: 10.3847/2041-8213/ac7fa3DOI
  7. [7]Bardalez Gagliuffi, D. C., Balmer, W. O., Pueyo, L., et al. 2025, ApJL, 988, L18, doi: 10.3847/2041-8213/ade30fDOI
  8. [8]Bendahan-West, R., Marino, S., Carter, A. L., et al. 2026, MNRAS, 546, staf2255, doi: 10.1093/mnras/staf2255
  9. [9]Biller, B., Carter, A. L., Kraus, A. L., et al. 2025, JWST Proposal. Cycle 3, ID. #6005
  10. [10]Bogat, E., Schlieder, J. E., Lawson, K. D., et al. 2025, AJ, 170, 225, doi: 10.3847/1538-3881/adfc75DOI
  11. [11]Bowens-Rubin, R., Mang, J., Limbach, M. A., et al. 2025, ApJL, 986, L26, doi: 10.3847/2041-8213/addbde
  12. [12]Bowler, B. P. 2016, PASP, 128, 102001, doi: 10.1088/1538-3873/128/968/102001DOI
  13. [13]Bradley, L., Sipőcz, B. M., Robitaille, T. P., et al. 2026, 3.1.0 Zenodo, doi: 10.5281/zenodo.23353738
  14. [14]Burn, R., Schlecker, M., Mordasini, C., et al. 2021, A&A, 656, A72, doi: 10.1051/0004-6361/202140390DOI
  15. [15]Burrows, A., Sudarsky, D., & Lunine, J. I. 2003, ApJ, 596, 587, doi: 10.1086/377709DOI
  16. [16]Carter, A. 2021, in European Planetary Science Congress, EPSC2021-415, doi: 10.5194/epsc2021-415
  17. [17]Carter, A., & Biller, B. 2026, JWST Proposal. Cycle 5, ID. #10764
  18. [19]Carter, A., Absil, O., Balmer, W., et al. 2024, JWST Proposal. Cycle 3, ID. #5835
  19. [20]Carter, A. L., Hinkley, S., Bonavita, M., et al. 2021, MNRAS, 501, 1999, doi: 10.1093/mnras/staa3579DOI
  20. [22]Chabrier, G., Baraffe, I., Phillips, M., & Debras, F. 2023, A&A, 671, A119, doi: 10.1051/0004-6361/202243832DOI
  21. [23]Chomez, A., Delorme, P., Lagrange, A.-M., et al. 2025, A&A, 697, A99, doi: 10.1051/0004-6361/202451751DOI
  22. [24]Crotts, K. A., Carter, A. L., Lawson, K., et al. 2025, ApJL, 987, L41, doi: 10.3847/2041-8213/ade798DOI
  23. [25]Crotts, K. A., Carter, A. L., Biller, B., et al. 2026, ApJ, 1008, 22, doi: 10.3847/1538-4357/ae9083DOI
  24. [26]Davis, C. E., Fortney, J. J., Iyer, A., et al. 2025, ApJ, 994, 198, doi: 10.3847/1538-4357/ae1015DOI
  25. [27]Desidera, S., Chauvin, G., Bonavita, M., et al. 2021, A&A, 651, A70, doi: 10.1051/0004-6361/202038806DOI
  26. [28]Emsenhuber, A., Mordasini, C., Burn, R., et al. 2021, A&A, 656, A70, doi: 10.1051/0004-6361/202038863DOI
  27. [29]Fernandes, R. B., Mulders, G. D., Pascucci, I., Mordasini, C., & Emsenhuber, A. 2019, ApJ, 874, 81, doi: 10.3847/1538-4357/ab0300DOI
  28. [30]Ferrer-Chavez, R., Wang, J. J., Wagner, K., et al. 2026, arXiv e-prints, arXiv:2609.20748, doi: 10.48550/arXiv.2609.20748
  29. [31]Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306, doi: 10.1086/670067
  30. [32]Fulton, B. J., Rosenthal, L. J., Hirsch, L. A., et al. 2021, ApJS, 255, 14, doi: 10.3847/1538-4365/abfcc1DOI
  31. [33]Gagné, J. 2024, PASP, 136, 063001, doi: 10.1088/1538-3873/ad4e6aDOI
  32. [34]Gagné, J., Mamajek, E. E., Malo, L., et al. 2018, ApJ, 856, 23, doi: 10.3847/1538-4357/aaaee09
  33. [35]Girard, J. H., Leisingring, J., Kammerer, J., et al. 2022, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 12180, Space Telescopes and Instrumentation 2022: Optical, Infrared, and Millimeter Wave, ed. L. E. Coyle, S. Matsuura, & M. D. Perrin, 121803Q, doi: 10.1117/12.2629636
  34. [36]Hinkley, S., Carter, A. L., Ray, S., et al. 2022, PASP, 134, 095003, doi: 10.1088/1538-3873/ac77bdDOI
  35. [37]Joye, W. A., & Mandel, E. 2003, in Astronomical Society of the Pacific Conference Series, Vol. 295, Astronomical Data Analysis Software and Systems XII, ed. H. E. Payne, R. I. Jedrzejewski, & R. N. Hook, 489
  36. [38]Kammerer, J., Girard, J., Carter, A. L., et al. 2022, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 12180, Space Telescopes and Instrumentation 2022: Optical, Infrared, and Millimeter Wave, ed. L. E. Coyle, S. Matsuura, & M. D. Perrin, 121803N, doi: 10.1117/12.2628865
  37. [39]Karalidi, T., Marley, M., Fortney, J. J., et al. 2021, ApJ, 923, 269, doi: 10.3847/1538-4357/ac3140DOI
  38. [40]Krist, J. E., Balasubramanian, K., Beichman, C. A., et al. 2009, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7440, Techniques and Instrumentation for Detection of Exoplanets IV, ed. S. B. Shaklan, 74400W, doi: 10.1117/12.826448
  39. [41]Krist, J. E., Balasubramanian, K., Muller, R. E., et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7731, Space Telescopes and Instrumentation 2010: Optical, Infrared, and Millimeter Wave, ed. J. M. Oschmann, Jr., M. C. Clampin, & H. A. MacEwen, 77313J, doi: 10.1117/12.856488
  40. [42]Lacy, B., & Burrows, A. 2023, ApJ, 950, 8, doi: 10.3847/1538-4357/acc8cbDOI
  41. [43]Lafrenière, D., Marois, C., Doyon, R., & Barman, T. 2009, ApJL, 694, L148, doi: 10.1088/0004-637X/694/2/L148DOI
  42. [44]Lagrange, A.-M., Wilkinson, C., Målin, M., et al. 2025, Nature, 642, 905, doi: 10.1038/s41586-025-09150-4DOI
  43. [45]Langlois, M., Gratton, R., Lagrange, A.-M., et al. 2021, A&A, 651, A71, doi: 10.1051/0004-6361/202039753DOI
  44. [46]Linder, E. F., Mordasini, C., Mollière, P., et al. 2019a, A&A, 623, A85, doi: 10.1051/0004-6361/201833873DOI
  45. [47]Linder, E. F., Mordasini, C., Mollière, P., et al. 2019b, A&A, 623, A85, doi: 10.1051/0004-6361/201833873DOI
  46. [48]Lodders, K., & Fegley, B. 2002, Icarus, 155, 393, doi: 10.1006/icar.2001.6740DOI
  47. [49]Mang, J., Chachan, Y., Morley, C. V., et al. 2026a, ApJ, 1008, 254, doi: 10.3847/1538-4357/ae9609DOI
  48. [51]Marleau, G.-D., Coleman, G. A. L., Leleu, A., & Mordasini, C. 2019, A&A, 624, A20, doi: 10.1051/0004-6361/201833597DOI
  49. [52]Marley, M. S., & Robinson, T. D. 2015, ARA&A, 53, 279, doi: 10.1146/annurev-astro-082214-122522DOI
  50. [53]Marley, M. S., Saumon, D., Visscher, C., et al. 2021, ApJ, 920, 85, doi: 10.3847/1538-4357/ac141dDOI
  51. [54]Matthews, E. C., Carter, A. L., Pathak, P., et al. 2024, Nature, 633, 789, doi: 10.1038/s41586-024-07837-8DOI
  52. [55]Mawet, D., Milli, J., Wahhaj, Z., et al. 2014, ApJ, 792, 97, doi: 10.1088/0004-637X/792/2/97DOI
  53. [56]Meyer, M. R., Li, Y., Calissendorff, P., & Amara, A. 2025, arXiv e-prints, arXiv:2508.05122, doi: 10.48550/arXiv.2508.05122DOI
  54. [57]Mignon, L., Delfosse, X., Meunier, N., et al. 2025, A&A, 700, A146, doi: 10.1051/0004-6361/202451142DOI
  55. [58]Morley, C. V., Marley, M. S., Fortney, J. J., et al. 2014, ApJ, 787, 78, doi: 10.1088/0004-637X/787/1/78DOI
  56. [59]Morley, C. V., Mukherjee, S., Marley, M. S., et al. 2024, ApJ, 975, 59, doi: 10.3847/1538-4357/ad71d5DOI
  57. [60]Mukherjee, S., Fortney, J. J., Morley, C. V., et al. 2024, ApJ, 963, 73, doi: 10.3847/1538-4357/ad18c2DOI
  58. [61]Nielsen, E. L., De Rosa, R. J., Macintosh, B., et al. 2019, AJ, 158, 13, doi: 10.3847/1538-3881/ab16e9DOI
  59. [62]Öberg, K. I., Murray-Clay, R., & Bergin, E. A. 2011, ApJL, 743, L16, doi: 10.1088/2041-8205/743/1/L16DOI
  60. [63]Palatnick, S., Millar-Blanchaer, M. A., Zhang, J., et al. 2025, ApJ, 995, 149, doi: 10.3847/1538-4357/ae1963DOI
  61. [64]Pathak, P., Petit dit de la Roche, D. J. M., Kasper, M., et al. 2021, A&A, 652, A121, doi: 10.1051/0004-6361/202140529DOI
  62. [65]Pearce, T. D., Launhardt, R., Ostermann, R., et al. 2022, A&A, 659, A135, doi: 10.1051/0004-6361/202142720DOI
  63. [66]Perrin, M. D., Sivaramakrishnan, A., Lajoie, C.-P., et al. 2014, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 9143, Space Telescopes and Instrumentation 2014: Optical, Infrared, and Millimeter Wave, ed. J. M. Oschmann, Jr., M. C. Clampin, G. G. Fazio, & H. A. MacEwen, 91433X, doi: 10.1117/12.2056689
  64. [67]Perrin, M. D., Soummer, R., Elliott, E. M., Lallo, M. D., & Sivaramakrishnan, A. 2012, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 8442, Space Telescopes and Instrumentation 2012: Optical, Infrared, and Millimeter Wave, ed. M. C. Clampin, G. G. Fazio, H. A. MacEwen, & J. M. Oschmann, Jr., 84423D, doi: 10.1117/12.925230
  65. [68]Phillips, M. W., Tremblin, P., Baraffe, I., et al. 2020a, A&A, 637, A38, doi: 10.1051/0004-6361/201937381DOI
  66. [69]Phillips, M. W., Tremblin, P., Baraffe, I., et al. 2020b, A&A, 637, A38, doi: 10.1051/0004-6361/201937381DOI
  67. [70]Pueyo, L. 2016, ApJ, 824, 117, doi: 10.3847/0004-637X/824/2/117DOI
  68. [71]Radcliffe, A., Charnay, B., Lagrange, A.-M., et al. 2026, A&A, 711, A257, doi: 10.1051/0004-6361/202659006DOI
  69. [72]Ruane, G., Ngo, H., Mawet, D., et al. 2019, AJ, 157, 118, doi: 10.3847/1538-3881/aa fee2
  70. [73]Saumon, D., & Marley, M. S. 2008, ApJ, 689, 1327, doi: 10.1086/592734DOI
  71. [74]Soderblom, D. R., Hillenbrand, L. A., Jeffries, R. D., Mamajek, E. E., & Naylor, T. 2014, in Protostars and Planets VI, ed. H. Beuther, R. S. Klessen, C. P. Dullemond, & T. Henning, 219–241, doi: 10.2458/azu.uapress_9780816531240-ch010
  72. [75]Soummer, R., Pueyo, L., & Larkin, J. 2012, ApJL, 755, L28, doi: 10.1088/2041-8205/755/2/L28DOI
  73. [76]Squicciarini, V., & Bonavita, M. 2022, A&A, 666, A15, doi: 10.1051/0004-6361/202244193DOI
  74. [77]Squicciarini, V., Mazoyer, J., Lagrange, A.-M., et al. 2025, A&A, 693, A54, doi: 10.1051/0004-6361/202452310DOI
  75. [78]Squicciarini, V., Desidera, S., Chauvin, G., et al. 2026, A&A, 711, A136, doi: 10.1051/0004-6361/202658983DOI
  76. [79]Sudarsky, D., Burrows, A., & Hubeny, I. 2003, ApJ, 588, 1121, doi: 10.1086/374331DOI
  77. [80]Suzuki, D., Bennett, D. P., Sumi, T., et al. 2016, ApJ, 833, 145, doi: 10.3847/1538-4357/833/2/145DOI
  78. [81]Thompson, W., Blakely, D., Xuan, J. W., et al. 2026, AJ, 172, 53, doi: 10.3847/1538-3881/ae64edDOI
  79. [82]Vigan, A., Fontanive, C., Meyer, M., et al. 2021, A&A, 651, A72, doi: 10.1051/0004-6361/202038107DOI
  80. [83]Wang, J. J., Ruffio, J.-B., De Rosa, R. J., et al. 2015, Astrophysics Source Code Library, record ascl:1506.001
  81. [84]Wittenmyer, R. A., Wang, S., Horner, J., et al. 2020, MNRAS, 492, 377, doi: 10.1093/mnras/stz3436DOI
  82. [85]Wogan, N. F., Mang, J., Batalha, N. E., et al. 2025, Research Notes of the American Astronomical Society, 9, 108, doi: 10.3847/2515-5172/add407DOI
  83. [86]Yelverton, B., Kennedy, G. M., Su, K. Y. L., & Wyatt, M. C. 2019, MNRAS, 488, 3588, doi: 10.1093/mnras/stz1927DOI
  84. [87]Zahnle, K. J., & Marley, M. S. 2014, ApJ, 797, 41, doi: 10.1088/0004-637X/797/1/41DOI
  85. [88]Zhang, S., Kalscheur, M., Long, F., et al. 2023, ApJ, 952, 108, doi: 10.3847/1538-4357/acd334DOI

Paper details

Contents