Constraining nearby substellar companion architectures using High Contrast Imaging, Radial Velocity and Astrometry L. F. Sartori ,, M. J. Bonse, Y. Li, F. A. Dannert, S. P. Quanz, C. Lovis, A. Boehle Abstract Context. Nearby stars offer prime opportunities for exoplanet discovery and characterization through various detection methods. By combining high-contrast imaging (HCI), radial velocity (RV), and astrometry, it is possible to better constrain the presence of substellar companions, as each method probes different regions of their parameter space. A detailed census of planets around nearby stars is essential to guide the selection of targets for future space missions seeking to identify Earth-like planets and potentially habitable worlds. In addition, the detection and characterisation of giant planets and brown dwarfs is crucial for understanding the formation and evolution of planetary systems. Aims. We aim to constrain the possible presence of substellar companions for 7 nearby M-dwarf stars (GJ 3325, GJ 1125, GJ 367, GJ 382, GJ 402, GJ 465 and GJ 357) using a combination of new SPHERE/ H2 HCI and archival RV and astrometric data. We investigate how combining these techniques improves the detection constraints for giant planets and brown dwarfs compared to using each method individually. Methods. For each star and each data set, we compute the mass limits as a function of semi-major axis or projected separation using standard techniques. We then use a Monte Carlo approach to assess the completeness of the companion mass / semi-major axis parameter space probed by the combination of the three methods, as well as by the three methods independently. Results. Our combined approach significantly increases the fraction of detectable companions. We quantify improvements of up to ∼ 60 % \sim 60\% over RV alone (improvement achieved for GJ 3325), ∼ 50 % \sim 50\% over HCI alone (for GJ 3325), and ∼ 12 % \sim 12\% over astrometry alone (for GJ 367). Although no new companion was detected, we could place stronger constraints on potential substellar companions. Conclusions. The combination of HCI, RV and astrometry provides significant improvements in the detection of substellar companions over a wider parameter space. Applying this approach to larger samples and lower-mass companions will help constraining the search space for future space missions aimed at finding potentially habitable or even inhabited planets. 1 Introduction In the coming years and decades, large scale exoplanet imaging surveys and space missions will search for exoplanets and brown dwarfs around hundreds of stars, with a focus on the Solar neighbourhood. These surveys will be performed with instruments such as the ERIS/NIX imager at the VLT ( Dubber et al. 2022 ; Davies et al. 2023 ), the Mid-infrared ELT Imager and Spectrograph (METIS, Quanz et al. 2015 ; Brandl et al. 2021 ) or the Planetary Camera and Spectrograph (PCS, Kasper et al. 2021 ) at the ELT, as well as by space missions such as the Large Interferometer for Exoplanets (LIFE, Quanz et al. 2018 ; Quanz et al. 2022 ) and the Habitable Worlds Observatory (HWO, The LUVOIR Team 2019 ; Gaudi et al. 2020 ; National Academies of Sciences & Medicine 2021 ). The goal of such efforts will be to determine the architecture and demographics of planetary systems, investigate planet formation and evolution, and ultimately look for life beyond the Solar System. In this context, constraining the architecture of substellar companions, particularly giant planets and brown dwarfs, is essential for advancing our understanding of planetary system formation and evolution (e.g., Vorobyov 2013 ; Kley 2017 ; Matsukoba et al. 2023 ). These objects, which bridge the gap between the most massive planets and the least massive stars, can form through distinct mechanisms such as core accretion (e.g., Goldreich & Ward 1973 ; Pollack et al. 1996 ; Ida & Lin 2004 ; Mordasini et al. 2009 ; Mordasini et al. 2012 ) or gravitational instability (e.g., Boss 1998 ; Vorobyov 2013 ; Mayer et al. 2002 ; Boss 2006 ), each resulting in characteristic mass and orbital distributions (e.g., Janson et al. 2011 ; Rameau et al. 2013 ; Nielsen et al. 2019 ). By characterising their statistics, both in terms of population number and orbital arrangement, we can gain crucial insights into the relative importance of these formation pathways. Additionally, the presence and configuration of such massive companions can significantly influence the dynamical evolution of planetary systems, affecting the stability and habitability of inner, terrestrial planets (e.g., Konopacky et al. 2016 ; Wang et al. 2018 ; Nielsen et al. 2020 ; Bitsch et al. 2015 ; Izidoro et al. 2015 ). Consequently, accurate constraints on the occurrence and orbital properties of giant planets and brown dwarfs in nearby systems are crucial not only for population studies but also for guiding future observational campaigns aimed at detecting potentially habitable exoplanets around these stars. Given the limited availability of observing time, the large-scale surveys and missions mentioned above will hardly be able to conduct blind searches for substellar companions. Therefore, stellar target samples must be carefully pre-selected based on current knowledge of planetary system statistics and architectures, as well as theoretical models of their formation and evolution (e.g., Dubber et al. 2022 ; Tuchow et al. 2024 ; Menti et al. 2024 ). This pre-selection is primarily informed by the characteristics of known exoplanetary systems and accepted formation models (e.g., Meyer et al. 2018 ; Nielsen et al. 2019 ; Vigan et al. 2021 ; Barbieri 2023 ). Several recent imaging programs have also demonstrated the effectiveness of using RV trends and/or high SNR astrometry to guide target selection, thereby increasing the likelihood of detecting massive companions (e.g., Calissendorff & Janson 2018 ; Nowak et al. 2020 ; Bonavita et al. 2022 ; Mesa et al. 2022 ; Mesa et al. 2023 ; Hinkley et al. 2023 ; Barbato et al. 2025 ). However, despite the thousands of companions discovered to date, the existing sample is highly biased due to the intrinsic limitations of the various detection methods (e.g., Perryman 2018 ). Radial velocity (RV) observations are most effective at detecting massive companions on close orbits, such as hot Jupiters, or smaller planets in even tighter configurations. High-contrast imaging (HCI) is sensitive to large, young, and hot planets on wide orbits. Astrometry, on the other hand, is particularly suited for detecting massive companions at intermediate orbital distances, especially around nearby stars. Each method therefore explores a distinct region of the companion mass and semi-major axis parameter space, which must be accounted for when characterising the demographics of planetary systems. By combining results from multiple detection techniques, a broader and more representative picture of the population of substellar companions can be obtained. In addition to this, existing archival data, which often remain unpublished due to non-detections, represent an underutilized resource. Properly analyzing these data to establish upper mass limits and detection thresholds can provide stringent constraints on the parameter space for potential substellar companions (e.g., Bonfils et al. 2013 ; Boehle et al. 2019 ; Hurt et al. 2021 ; Bonse et al. 2024 ; Dietrich 2024 ). These non-detections are valuable for delineating regions of parameter space that could or cannot host (yet) undetected companions, thereby refining our understanding of the true architecture of nearby planetary systems. The combination of complementary detection techniques to achieve a more comprehensive understanding of planetary and stellar systems has been explored in recent studies (e.g., Boehle et al. 2019 ; Dulz et al. 2020 ; Lagrange et al. 2020 ; Wood et al. 2021 ; Dietrich 2024 ; Gratton et al. 2024 ; Gratton et al. 2025 ; Barbato et al. 2025 ). For instance, Boehle et al. 2019 combined HCI observations from VLT/NACO ( Rousset et al. 2003 ; Lenzen et al. 2003 ) with archival RV data for six nearby stars, showing that HCI can increase the completeness for certain masses and semi-major axes up to 99\% compared to RV alone. Expanding on this approach, Wood et al. 2021 incorporated Gaia astrometric data, utilizing the renormalised unit weight error (RUWE; Lindegren et al. 2018 ) to constrain the properties of potential stellar companions in binary systems. These studies highlight the utility of multi-technique analyses in refining our constraints on the architectures of nearby planetary systems, by leveraging the strengths of each detection method and mitigating their individual limitations. In this study, we combine newly obtained HCI observations of 7 nearby stars acquired with the Spectro-Polarimetric High-contrast Imager for Exoplanet REsearch (SPHERE; Beuzit et al. 2019 ; Vigan et al. 2010 ) with archival RV data and proper motion anomalies (PMa) derived from the Hipparcos ( Perryman et al. 1997 ; van Leeuwen 2007 ) and Gaia (E)DR3 ( Gaia Collaboration et al. 2016 ; Gaia Collaboration 2020 ; Gaia Collaboration et al. 2021a ; Gaia Collaboration et al. 2021b ) catalogues ( Kervella et al. 2021 ; Kervella et al. 2022 ). While the renormalised unit weight error (RUWE) parameter is available for all Gaia sources and provides a broad indicator of astrometric anomalies, the PMa offers more direct and quantitative constraints on the properties of potential companions, making it particularly suitable for this study. This work is conducted as part of the Planetary Systems In Our Neighbourhood (PSION) project, which aims to identify and characterise companions around some of the closest stars to the Sun, in order to constrain the architectures of their planetary systems and to prioritise targets for future observational campaigns (e.g., Boehle et al. 2019 ; Sartori et al. 2023 ; Menti et al. 2024 ). Although the present results are limited by the small sample size, this study demonstrates the effectiveness of combining HCI, RV, and astrometric data for constraining the parameter space of (substellar) companions. It can also can be regarded as a proof-of-concept for future more comprehensive studies as the methodology can be easily applied to larger samples, providing valuable input for defining target lists for upcoming space missions such as LIFE and the HWO. The paper is organised as follows. In Section 2 we describe the sample selection, the available data (new and archival) and the performed data reduction. In Section 3 we illustrate the analysis performed to compute the constraints for each technique alone, and how we combine them to investigate the architecture of the systems. In Section 4 we summarise the results and compare them with the expectation from population models. Finally, we discuss our findings in Section 5 . 2 Sample and data 2.1 Target stars Our sample consists of seven nearby ( d<10 pc) M dwarf stars for which we obtained new high-contrast imaging in the H2 and H3 bands with the SPHERE instrument (Program ID 0104.C - 0336, PI A. Boehle, and Program ID 112.25G2, PI L. Sartori. See Section 2.2 for details): GJ 3325, GJ 1125, GJ 367, GJ 382, GJ 402, GJ 465 and GJ 357. Details about these stars are summarised in Table 1 . Since our primary goal is to demonstrate the potential of the combined method rather than to perform a detailed stellar characterization, we adopt literature values for stellar parameters where available, acknowledging the resulting inhomogeneities. As part of the PSION project, the targets were chosen from a volume-limited sample of all known single, dwarf stars within 10 pc with spectral types from F 0 to M 4 (94 stars in total). To compile the total sample, we first compared the predicted detection limits in the companion mass/semi-major axis parameter space from coronagraphic IRDIS H2/H3 dual band imaging 1 1 1 Based on the ESO SPHERE/IRDIS ETC calculator https://www.eso.org/observing/etc/ to those derived from archival long-term radial velocity data, mostly from the High Accuracy Radial velocity Planet Searcher (HARPS, Pepe et al. 2002 ; Mayor et al. 2003 ) and High Resolution Echelle Spectrometer (HIRES, Vogt et al. 1994 ) instruments (see Section 2.3 ), and selected the stars for which the SPHERE observations were predicted to be sensitive enough to substellar companions that are currently missed by the radial velocity data. Specifically, we predicted the HCI contrast curves using the ESO Exposure Time Calculator, while for the RV limits we followed the same procedure as described in Section 3.2 . We then checked that their apparent magnitude is bright enough to enable the use of the wavefront sensor of the SPHERE extreme AO system ( r<11 mag, Fusco et al. 2006 ). The final sample consisted of the aforementioned 7 M dwarf stars with no previous archival SPHERE observations. We obtained the requested observations for 5 of them during the ESO period P104, while 3 were observed during the ESO period P112. Among the stars in our sample, only GJ 367 and GJ 357 have previously detected exoplanet companions. Recent work based on TESS and HARPS data for GJ 367 ( Lam et al. 2021 ; Goffo et al. 2023 ) confirmed the presence of three planets with masses m {}_{\rm b}=0.6\pm 0.083 M ⊕ , m {}_{\rm c}=4.08\pm 0.30 M ⊕ and m {}_{\rm d}=5.93\pm 0.45 M ⊕ , and orbiting periods P {}_{\rm b}=0.3219225\pm 0.0000002 days, P {}_{\rm c}=11.543\pm 0.005 days and P {}_{\rm d}=34.39\pm 0.06 days. These planets are detected also in our archival RV timeseries, and we discuss in Section 2.3 how we treated the corresponding signal in our analysis. For GJ 357, the analysis of TESS and CARMENES data ( Luque et al. 2019 ) confirmed the presence of three planets with masses m {}_{\rm b}=1.84\pm 0.031 M ⊕ , m {}_{\rm c}=3.40\pm 0.46 M ⊕ and m {}_{\rm d}=6.1\pm 1.0 M ⊕ , and orbiting periods P {}_{\rm b}=3.93 days, P {}_{\rm c}=9.12 days and P {}_{\rm d}=55.7 days. We have treated these data in the same way as described for GJ 367. Because of the short periods and small semi-major axes, none of these planets is visible in our HCI data. Target Spectral Type Distance Stellar Mass Stellar Age T eff V References [pc] [M ⊙ ] [Gyr] [K] [mag] GJ 3325 M3V 9.23 0.290 \pm 0.014 5.6 3401 11.73 [1],[2],[3],[4],[5] GJ 1125 M3.5V 9.90 0.303 4.98 3089 11.71 [4],[6],[7],[8],[9] GJ 367 M1.0 9.41 0.455 8.65 3283 9.98 [4],[6],[8],[10] GJ 382 M2V 7.70 0.508 5.6 3359 9.26 [3], [6],[8],[11],[12] GJ 402 M4V 6.97 0.282 \pm 0.014 7.5 3001 11.67 [2],[6],[11],[13],[14] GJ 465 dM2.0 8.88 0.274 \pm 0.014 1.82 3219 11.27 [2],[4],[6],[9],[15] GJ 357 M2.5V 9.44 0.346 \pm 0.20 3204 10.91 [6],[4],[8],[9],[16] Table 1: Stellar properties of the seven targets of this paper. References : Spectral Type: [1] Henry et al. 2002 ; [4] Koen et al. 2010 ; [7] Lépine et al. 2013 ; [11] Kirkpatrick et al. 1991 ; [15] Schweitzer et al. 2019 ; [16] Gray et al. 2006 ; Stellar Mass: [2] Winters et al. 2021 ; [8] Stassun et al. 2019 ; Stellar Age: [3] Yee et al. 2017 ; [9] Queiroz et al. 2023 ; [10] Lu et al. 2024 ; [13] Brandt & Huang 2015 ; Effective Temperature: [5] Maldonado et al. 2020 ; [6] Gaia Collaboration et al. 2022 ; V mag: [4] Koen et al. 2010 ; [12] Kiraga 2012 ; [14] Landolt 1992 2.2 High-contrast imaging We obtained SPHERE high contrast imaging data of the seven target stars between January and March 2020, and between January and March 2024. Observations are taken in the IRDIFS mode, which employs simultaneously the infrared dual-band imager and spectrograph (IRDIS, Dohlen et al. 2008 ) in its dual-band imaging mode (DBI, Vigan et al. 2010 ) and the integral field spectrocopic mode (IFS, Claudi et al. 2008 ; Mesa et al. 2015 ). For DBI we used the H2H3 filter pair ( \lambda_{H2}=1.593\penalty\ \mu m, \Delta\lambda_{H2}=0.052\penalty\ \mu m, \lambda_{H3}=1.667\penalty\ \mu m, \Delta\lambda_{H3}=0.056\penalty\ \mu m) in pupil-stabilized mode to allow angular differential imaging (ADI, Marois et al. 2006 ). This filter pair is ideal for searching for planets and eventually studying their atmospheres, as it covers one of the main methane absorption features (e.g., Bonnefoy et al. 2018 ). For simultaneous IFS observations, we used the YJ mode ( \lambda=0.95-1.35\mu m, spectral resolution R\sim 54 ), but these data are not discussed in this paper, as they do not add additional information useful for our analysis (no planet is detected). The observations were taken with the ALC \_ YJH \_ S Lyot coronagraph ( Soummer 2005 ) which is optimised for the H -band and has an inner working angle IWA \sim 0".15 2 2 2 From the SPHERE User Manual, https://www.eso.org/sci/facilities/paranal/instruments/sphere/doc.html . To obtain a model of the point spread function (PSF), the stars were observed before and after the science sequence without the coronagraph but with an additional neutral density filter, ND2.0, with a transmission of \sim 1\% ( 1.2\% for H2 and 1.7\% for H3 ) of the total bandwidth 3 3 3 For GJ382, the neutral density filter ND3.5 was used instead of the ND2.0 one. . Additional information about the observations are provided in Table 2 . We obtained the raw files from the ESO archive and performed a standard reduction using the vlt-sphere package ( Vigan 2020 ), which relies on the official ESO pipeline ( ESO CPL Development Team 2014 ). The standard calibration includes dark and background subtraction, flat field and bad pixels correction, wavelength calibration (for the IFS data) and star centering. We then used the state-of-the-art pipeline for direct-imaging data reduction and analysis PynPoint ( Stolker et al. 2019 ) to divide the reduced DBI data cubes into single bands, median combine the point spread function (PSF) cubes obtained before and after the science observations, as well as perform Angular Differential Imaging (ADI) and Principal Component Analysis (PCA) ( Amara & Quanz 2012 , Soummer et al. 2012 ) PSF subtraction. The residual images for H2 are shown in Fig. 10 . No companion signal is detected in these images. We could therefore compute the contrast curves without subtracting any signal, as described in Section 3.1 . Target UT date FoV rot. Seeing (a) [deg] [arcsec] GJ 3325 2020-01-19 78.5 0.50\pm 0.04 GJ 1125 2020-02-07 31.9 0.71\pm 0.22 GJ 367 2020-02-08 30.4 0.64\pm 0.13 GJ 382 2020-02-22 37.3 0.76\pm 0.10 GJ 402 2024-01-21 25.5 0.33\pm 0.05 GJ 465 2024-03-07 94.2 0.54\pm 0.06 GJ 357 2024-01-20 126.5 0.47\pm 0.08 IRDIS filter H2H3 IFS band YJ IRDIS DIT (b) \times NDIT (c) 8\times 2 s IRDIS T exp 46.9 min IFS DIT \times NDIT 32\times 1 s IFS T exp 51.7 min Table 2: Summary of the new SPHERE observations of the seven targets. (a) Mean seeing during observations. (b) Detector Integration Time in seconds. (c) Number of frames per dithering position. 2.3 Radial velocity The archival HARPS and HIRES radial velocity measurements used in this work are summarised in Table 4 . The data were processed and analysed within the Data & Analysis Center for Exoplanets (DACE) platform, a facility hosted by the University of Geneva which is dedicated to extrasolar planets data visualisation, exchange and analysis 4 4 4 https://dace.unige.ch/dashboard/index.html . The RV and stellar activity indicators time series available in DACE are computed by cross-correlating the calibrated spectra with stellar templates of the corresponding stellar class using state-of-the-art pipelines 5 5 5 Specifically, the RVs and activity indicators in DACE are derived using the Data Reduction Software (DRS) for HARPS, which employs the cross-correlation function (CCF) method ( Pepe et al. 2002 ; Mayor et al. 2003 ), and the Keck Observatory HIRES pipeline for HIRES, which also uses the CCF method ( Vogt et al. 1994 ; Isaacson & Fischer 2010 ). . We retrieved the available RV time series using the DACE python APIs, and analysed them in a similar way as described in Sartori et al. 2023 . In summary, for each target we subtracted a linear offset arising from the recession velocity of the star itself and from instruments cross-calibration, added the instrumental jitter to the error bars, and then computed the generalised Lomb-Scargle periodograms ( Cumming et al. 2008 , with formalism from Zechmeister & Kürster 2009 ). For all stars except GJ 367 and GJ 357, the obtained periodograms do not show any peak indicating the presence of an orbiting planet at any of the probed periods. We therefore used the obtained RV time series without subtracting any signal to compute the mass limits, as described in Section 3.2 . For GJ 367 and and GJ 357 instead we were able to reproduce the analysis in the discovery papers our Keplerian fit returned values consistent with those in Goffo et al. 2023 Luque et al. 2019 , respectively. We subtracted the signal arising from the known planets before computing the mass limits. We note that in order to ensure a consistent analysis across our sample and to avoid the need for re-reducing or homogenizing heterogeneous datasets, we limit our RV analysis to time series available through the DACE platform. While this excludes some additional measurements available in the literature (e.g., CARMENES, PFS, or UVES for GJ 357), we consider the available data sufficient for the illustrative purpose of this work. Target No. of measurements \Delta t_{\rm{RV}} (a) a for \Delta t_{\rm{RV}} (b) \sigma_{\rm{RV}} HIRES HARPS 03 HARPS 15 Total [days] [AU] [ms -1 ] GJ 1125 19 0 0 19 4090 3.4 17.3 GJ 367 0 25 346 371 6391 5.2 3.12 GJ 382 96 34 0 130 6068 5.2 4.3 GJ 3325 21 9 1 31 6961 4.7 6.2 GJ 402 24 6 5 45 5864 4.2 7.6 GJ 465 0 23 20 43 5140 3.8 6.2 GJ 357 61 53 0 114 5884 4.5 4.4 Table 3: Summary of the archival radial velocity (RV) observations available for four of the five targets in our sample. (a) Total time baseline for the RV measurements. (b) Semi-major axis a correspondent to a revolution with period equal to the total time baseline \Delta t_{\rm{RV}} , given the stellar masses in Table 1 . 2.4 Astrometry Seven of our targets are present in both the Hipparcos ( Perryman et al. 1997 ; van Leeuwen 2007 ) and Gaia EDR3 ( Gaia Collaboration et al. 2016 ; Gaia Collaboration 2020 ; Gaia Collaboration et al. 2021a ; Gaia Collaboration et al. 2021b ) catalogues. For this study we directly retrieved the proper motion anomaly (PMa) values from Kervella et al. 2021 . Details about the astrometric data are given in Table 4 . We have verified that the observations are not saturated (G mag >>3 ) and that all Gaia (E)DR3 measurements have low RUWE ( <1.2 for the entire sample) and low excess noise ( \epsilon_{i}<0.35 ) values, meaning that the measurements are reliable for our analysis and do not indicate the presence of stellar companions orbiting at large distances. Target GAIA DR3 ID HIPPARCOS ID \Delta v_{{\rm{T,GDR3}}} (a) r_{\rm{orb}} for \delta t_{\rm{GDR3}} (b) [ms -1 ] [AU] GJ 1125 3840711729206820000 46655 18.7\pm 7.3 1.36 GJ 367 5412250540681250560 47780 21.13\pm 5.55 1.56 GJ 382 3828238392559860992 49986 3.78\pm 2.33 1.62 GJ 3325 2976850598890749824 23512 1.2\pm 3.5 1.31 GJ 402 3864615459886222464 53020 2.9\pm 4.1 1.309 GJ 465 3519785523672576384 60559 2.1\pm 4.0 1.29 GJ 357 5664814198431308288 47103 1.6\pm 3.2 1.42 Table 4: Summary of the astrometric measurements from Kervella et al. 2021 for the four targets present in both the Hipparcos and Gaia EDR3 catalogues. (a) Norm of the tangential proper motion anomaly (PMa) vector converted to linear velocity using the Gaia EDR3 parallax. (b) Radius of an orbit with period corresponding to the Gaia DR3 observing window \delta t_{\rm{GDR3}}=1038 d given the stellar mass of the target. 3 Analysis 3.1 Contrast curves and mass limits from HCI We computed the analytical contrast curves for each star and for both the H2 and H3 filters separately using the python package applefy 6 6 6 https://applefy.readthedocs.io/ Bonse et al. 2023 . To account for the attenuation of the signal due to post-processing with PCA, we inserted artificial planets into the data using the unsaturated off-axis stellar PSF. The artificial planets were inserted at different separations every 1\lambda/D and at 6 equally spaced position angles. We inserted only one artificial planet at a time and ran our PSF subtraction for a range of PCA components \in\left[5,10,20,30,40,50,75,100\right] . Based on these extensive artificial planet experiments, we computed the contrast curves based on a t-test ( Mawet et al. 2014 ) , assuming that the residual noise was sufficiently Gaussian distributed. We set the detection threshold to a false positive fraction (FPF) of 2.87\times 10^{-7} , which is equivalent to 5\sigma at large separations from the star. The samples for the t-test were based on aperture sums as originally suggested by Mawet et al. 2014 . We accounted for the effect of the aperture placement by considering 20 different aperture positions for each test (see Bonse et al. 2023 ). The final contrast curves reach out to 5 arcsec (the maximum apparent separation allowed by our data) and are shown in Fig. 1 . We further converted the obtained contrast curves to mass limits assuming the stellar ages in Table 1 and the AMES-Cond evolutionary models ( Baraffe et al. 2003 ) . For stars without literature values, we assumed a stellar age of 5\pm 0.5 Gyr. The resulting mass limits are also shown in Fig. 1 . Figure 1: Left: Contrast curves computed from the SPHERE H2 high contrast images. The rise in most contrast curves at around 1 arcsec (converted to physical scale) corresponds to the limit of the AO correction (0.8 arcsec, Fusco et al. 2006 ; Fusco et al. 2016 ). The reached contrast in the H2 band and H3 band (not plotted here) is almost identical. Right: Mass limits derived from the SPHERE H2 contrast curves in above using the AMES-Cond evolutionary models ( Baraffe et al. 2003 ). The shaded regions show the possible spread due to the uncertainty on the stellar ages. We note that the mass limits obtained in the H2 filter are a factor \sim 2 better than in the H3 filter. 3.2 Mass limits from radial velocity As discussed in Section 2.3 , apart from GJ 367 the archival RV data do not show any evidence of companions orbiting the target stars. We can however use the RV timeseries to compute the mass limits for potential planets around these stars, i.e. to constrain the companion mass / semi-major axis parameter space where potential planets could still be present, but whose RV signal would be too weak to be detected with the currently available data. The mass limits as a function of both orbital period P and semi-major axis a are shown in Fig. 2 . Because of geometry, the companion mass m p is degenerate with the companion’s orbit inclination i , so that only a minimum mass m {}_{p}\sin(i) can be computed. To obtain the mass limits based on the RV observations we followed the procedure proposed in Bonfils et al. 2013 . In a first step, we created 1000 time series by randomly shuffling the original RV timeseries, computed the correspondent Lomb-Scargle periodograms as in Section 2.3 , and found the maximum power of each periodogram, pow max,i . The power that divides the lower 99\% from the upper 1\% of the values in \{ pow {}_{\rm{max},1},..., pow {}_{\rm{max},1000}\} represents the 1\% false alarm probability (FAP), which will be used in the following of the analysis. For each period P_{\rm{GLS}} considered in the original Lomb-Scargle periodogram we then simulated 12 RV time series by adding to the original RV data a sinusoidal with period P_{\rm{GLS}} , an initial semi-amplitude K_{\rm{in}}=0.5 m s -1 , and one of 12 equi-spaced phases T . Finally, for each period P_{\rm{GLS}} and phase T we increased K_{\rm{in}} until the correspondent signal in the Lomb-Scargle periodogram reached the 1\% FAP defined above, and converted the obtained K to m {}_{p}\sin(i) assuming Keplerian orbits and the stellar mass listed in Table 1 . For each period P_{\rm{GLS}} the mass limit corresponds to the average over the 12 trial phases T . The obtained mass limits for the 7 target stars with available archival RV data are shown in Fig. 2 . In the plots we also highlight the periods correspondent to the habitable zone (HZ) which we computed using the Bolometric correction from Habets & Heintze 1981 and the scaling values for the inner and outer radius from Kasting et al. 1993 , Kasting 1996 and Whitmire & Reynolds 1996 . We note that the available data would allow to detect planets in the HZ with minumum mass m {}_{p}\sin(i)=3-30 M ⊕ , depending on the target. Figure 2: Mass limits computed from the archival RV data using a bootstrap method (straight lines, following Bonfils et al. 2013 and references therein). The shadowed regions indicate the habitable zones (HZ, same color coding as for the mass limits). The targets have been split into two plots for ease of reading. The available RV data would allow us to detect planets in the HZ with minumum mass m {}_{p}\penalty\ \sin(i)\penalty\ =\penalty\ 1-30\penalty\ M ⊙ , depending on the target. Everything above the mass limits should have been detected. See Section 3.2 for details. 3.3 Mass constraints from astrometry The PMa is defined as the difference between the long-term proper motion (PM) vector measured between two astrometric surveys, in our case Hipparcos and Gaia EDR3, and the instantaneous PM vector measured from one of the two catalogues. This quantity provides an additional constraint on the mass and semi-major axis of potential companions orbiting our target stars, therefore influencing its motion around the common barycenter and the displacement of its photocenter. We computed these constraints following the procedure described in Kervella et al. 2019a ; Kervella et al. 2022 and the PMa measurements in the associated catalogue ( Kervella et al. 2021 ). Specifically, for the simplest case of m {}_{p}\ll M ⋆ and a circular orbit perpendicular to the line-of-sight, the relation between the companion mass m p , the stellar mass M ⋆ and the semi-major axis (orbital radius) r is given by: \frac{{\rm{m}}_{p}}{\sqrt{r}}=\sqrt{\frac{{\rm{M}}_{\star}}{G}}v_{1}=\sqrt{\frac{{\rm{M}}_{\star}}{G}}\left(\frac{\Delta\mu\text{ [mas yr${}^{-1}$]}}{\varpi\text{ [mas]}}\times 4740.47\right) (1) where v_{1} is the tangential orbital velocity of the primary star, G the Gravitational constant, \Delta\mu the PMa and \varpi the parallax (see Eq. 7 in Kervella et al. 2019a ). From Eq. 1 we obtain the companion mass sensitivity curve m {}_{p}(r)=f(r) . In a more realistic case, we need to take into account the uncertainty in the orbital inclination as well as observing window smearing and efficiency as a function of orbital period (see Section 3.6 in Kervella et al. 2019a for details about statistical and systematic corrections), so that the normalised companion mass sensitivity curve is given by: {\rm{m}}_{p}(r)=\frac{\sqrt{r}}{\gamma[P(r)/\delta t_{{\rm{GDR3}}}]}\sqrt{\frac{{\rm{M}}_{\star}}{G}}\frac{\Delta v_{{\rm{T,GDR3}}}}{\eta\zeta} (2) where P(r) is the period corresponding to the orbital radius r (semi-major axis), \Delta v_{{\rm{T,GDR3}}} is the norm of the tangential PMa vector converted to linear velocity using the Gaia DR3 parallax (from Kervella et al. 2021 ), \gamma and \zeta are two functions quantifying the window smearing and the efficiency as defined in Kervella et al. 2019a , \eta=87^{+12}_{-32}\% is a normalisation factor which accounts for the uncertainty in the orbital inclination ( Kervella et al. 2019a ), and \delta t_{\rm{GDR3}}=1038 d is the observing window for the Gaia DR3 mission ( Gaia Collaboration et al. 2021a ). The PMa sensitivity limits, i.e. the constraints on companion mass m p and semi-major axis a , which we computed for the seven targets present in both the Hipparcos and Gaia EDR3 catalogues using Eq. 2 and the PMa values in Kervella et al. 2021 , are shown in Fig. 3 . We note that these curves can be interpreted as a sort of mass limits. Indeed, the curves describe the possible (m_{p},r) combinations, given the known stellar mass M_{\star} and the observed tangential orbital velocity v_{1} . By keeping M_{\star} fixed, every planet with (m_{p},r) lying above the derived curve would have a tangential velocity v>v_{1} , and this would have been detected. Similarly, everything lying below would have v2 arcsec. In Fig. 1 (3rd and 4th rows) we show the same contrast curves converted to mass limits using the AMES-Cond evolutionary models ( Baraffe et al. 2003 ) for the given stellar ages, distances and apparent magnitudes. Our H2 images allow the detection of companions with masses down to 5-10 M J in the background-limited regions, and a factor \sim 2 higher in the H3 band. This difference is due to the fact that our observations probe the temperature regime T<1000 K where methane absorption is strong, so the flux in the methane absorption band is very weak (compared to the continuum). As the H2 filter provides better limits compared to H3 , we will concentrate on these observations only. We note that for two targets the stellar age is not known, so we assumed 5\pm 0.5 Gyr, consistent with other work in the PSION project. We have tested that for these targets, and for the other targets where the age is known, varying the age within the (assumed) uncertainties does not significantly change the results. The mass limits from RV are shown in Fig. 2 . The archival observations allow us to compute limits up to \sim 10^{4} days, corresponding to semi-major axis of \sim 10 AU. As expected, the limiting minimum mass m {}_{p}\sin(i) shows a power-law increase with period. These observations would allow to detect planets in the HZ with minimum masses m {}_{p}\penalty\ \sin(i)\penalty\ =\penalty\ 1-30\penalty\ M ⊙ . The companion mass sensitivity limit from PMa (astrometry) is shown in Fig. 3 , extending up to \sim 200 AU. Unlike RV, these curves already take into account the unknown inclination i , so the y-axis represents the companion mass itself and not a minimum mass. Due to observational window smearing and efficiency, the sensitivity at separations below the Gaia DR3 observing window \delta t_{\rm{GDR}}=1038 d ( 1-2 AU depending on the source) shows strong peaks and is difficult to interpret (at least in the context of this study). At larger separation, the sensitivity flattens between \sim 2-10 AU reaching a minimum companion mass of 0.1-1 M J depending on the source ( 6\times 10^{-2} for GJ 357), and shows a power-law increases at larger separations. These results clearly show how the different detection methods are sensitive to very different parts of parameter space. This is also well illustrated in Fig. 5 , which shows, for each target, the fraction of simulated planets that would be detected by each technique as a function of semi-major axis up to 100 AU. Specifically, RV and astrometry would detect the majority (up to \sim 100\% ) of close-in planets, and the detection fraction decreases rapidly with increasing semi-major axis, although the decrease is less pronounced for astrometry. The RV detection fraction indeed drops to \sim 40\% at 20-40 AU for most sources, while the astrometry detection fraction remains above \sim 80\% out to \sim 60 AU for 5 out of 7 targets. On the other hand, HCI misses up to \sim 60\% of the simulated companions at a<10 AU. The detection rate then slowly reaches a \sim 80\% “plateau” between \sim 20-40 AU, and drops steeply after that. 4.2 Architecture constraints from combining methods Fig. 4 and Fig. 11 - 16 show the percentage of the 1000 simulated companions for each mass/semi-major axis combination that could be detected by RV, HCI and astrometry, and by combinations of these methods. We are considering semi-major axes up to 2 times the SPHERE FOV, as companions orbiting at those separations can still have projected separations within the FOV for some orbital phases. Similar to what we noted for Fig. 5 , the limiting mass for RV and astrometry increases quickly as a function of semi-major axis, whereas the constraints from HCI are U-shaped with a flat plateau corresponding to the background limited-region and a sharp decrease at shorter separations as well as beyond the SPHERE FOV. Looking at the methods individually, for 5 of the 7 targets the best performance is provided by astrometry, which would be able to detect up to \sim 81-94\% of the simulated companions. For GJ 1125 and GJ 367, the best performance is provided by HCI, but with a lower percentage of detectable companions ( \sim 40\% ). The worst performance is always provided by RV, with \sim 15\% of detectable companions for the most distant star GJ 1125, and \sim 35-44\% of detectable companions for the other targets. This result is not surprising as RV is mostly sensitive to short period companions and is therefore not optimal at the largest semi-major axes considered in this study (see also target selection in Section 2.1 ). Combining all three methods, the percentage of detectable targets ranges from \sim 50\% to \sim 94\% among the 7 targets. Again, in 5 cases, astrometry provides the greatest enhancement compared to the detection fraction reached by the other two techniques, with \sim 34-43\% of companions detected exclusively by this technique. These companions usually lie in the lowest mass / shortest semi-major axis or highest mass / middle to largest semi-major axis parts of the considered parameter space, as these are not accessible to the other two techniques. For GJ 1125 and GJ 367 the greatest enhancement, is provided by HCI, with \sim 9-14\% of the simulated companions detected only by this technique, in agreement with Boehle et al. 2019 . These companions lie in a triangular shape, bounded at the bottom by the HCI mass limits in the background-limited regime and at the side by the RV and/or astrometry mass limits. 4.3 Survey sensitivity and companion population modeling Despite our combined analysis allowing us to search a rather large parameter space, we did not detect any new companion. This is not surprising given the known statistics of large planets and brown dwarfs. As discussed in Boehle et al. 2019 , HCI studies show that giant planets on a wide orbit are quite rare. For example, Bowler 2016 analysed a sample of FGKM stars and found that <3.9\% of M dwarfs host giant planets. Similarly, Meyer et al. 2018 applied combined statistic on a sample of M stars and obtained a planet detection rate of 1.5-3\% in the parameter range 106 stars) reaches 50-60 AU for companions with masses >10 M Jup . Similarly, the 2D survey (mean) completeness (Fig. 7 ) is the average of the individual detection maps. Again for the combined analysis the sensitivity is >95\% for companions with masses >10 M Jup up to separations a=50-60 AU. Figure 6: Depth of search for the three detection methods separately (top) and combined (bottom). The 2D survey depth (or depth of search of the survey) gives the number of stars around which the survey is sensitive for a given companion mass and semi-major axis. See Section 4.3 for details. Figure 7: Survey mean completeness for the three detection methods separately (top) and combined (bottom). The 2D survey (mean) completeness is computed as the average of the detection maps obtained for the individual targets. See Section 4.3 for details. To compare our survey results with theoretical expectations, we adopt a phenomenological model based on Meyer et al. (in prep.). This parametric model simultaneously fits the frequencies of both very low-mass brown dwarfs formed through “multiple-like processes” (fragmentation during early cloud collapse or disk fragmentation analogous to binary or multiple star formation), and gas giants formed via “planet-like processes”, such as core accretion or disk instability within a protoplanetary disk. These frequencies are modelled as functions of orbital separation and host star mass. The frequency data is derived from 50 independent point estimates compiled over 20 years of direct imaging, microlensing, and radial velocity surveys (e.g. Wittenmyer et al. 2016 , Borgniet et al. 2017 , Nielsen et al. 2019 , Vigan et al. 2021 , Fulton et al. 2021 , Grandjean et al. 2022 ). The parameter space encompasses companion masses ( m_{c}\sim 1-85 M J ), orbital separations ( a\sim 1-6000 AU), and stellar masses ( M_{*}\sim 0.2-3.0 M ⊙ ). We assume log-normal orbital separations and power-law mass ratio distributions for both planets and brown dwarfs around M dwarfs. The brown dwarf orbital distribution is fixed to the values in Winters et al. 2019 ( \mu_{BD}=1.43 , \sigma_{BD}=1.21 ) , with the projected separation (20 AU) adjusted to the physical separation (27 AU) using a correction factor of 1.35 from Duquennoy & Mayor 1991 . The final model is characterized by six parameters: two normalization factors for giant planets (GP) and brown dwarfs (BD) ( A_{GP} , A_{BD} ), two parameters for the log-normal semi-major axis distribution of gas giants ( \mu_{GP} , \sigma_{GP} ), and two power-law indices for the mass ratio distributions of giant planets and brown dwarfs ( \alpha_{GP} , \beta_{BD} ). We found ln (A_{GP})=-5.52_{-0.83}^{+0.93} , ln (\sigma_{GP})=0.53_{-0.07}^{+0.06} , ln (\mu_{GP})=-1.32_{-0.22}^{+0.18} , \alpha_{GP}=1.43_{-0.13}^{+0.12} , ln (A_{BD})=-3.78_{-0.47}^{+0.70} , and \beta_{BD}=0.36_{-0.34}^{+0.30} . These results show that giant planets peak at orbital distances of 4 AU and taper off beyond the 1-10 AU range, consistent with the latest exoplanet survey findings (e.g. Cumming et al. 2008 , Fulton et al. 2021 , Meyer et al. 2018 ). For our program, we model companions around M dwarfs as two distinct populations: (1) Giant Planets (0.03M J –0.1 M ⊙ ), and (2) Brown Dwarfs (3–75 M J ). The lower mass limit for giant planets is set at 0.03 M J (or 10 Earth masses), while the upper limit is defined as 10% of the stellar mass (0.1 M ⊙ ), rather than the conventional deuterium-burning threshold at 13 M J , to better reflect the maximum mass typically expected from disk-based formation processes. For planetary masses below 1 M J , we employ a custom sub-population model to describe so-called sub-Jovian planets, motivated by both the architecture of the Solar System and recent mm-wave ALMA disk continuum observations that show evidence of disk gaps possibly caused by embedded planets less than a few 10 M_{\oplus} ( van Boekel et al. 2017 ). For brown dwarfs, the lower mass limit is determined by the opacity limit for disk fragmentation, while the upper bound is set by the hydrogen-burning threshold of 75 M J . The mass/separation distribution models are defined as follows. For giant planets and brown dwarfs, we adopt the log-normal orbital distributions (Fig. 8 ) and power-law mass ratio distributions from the Meyer et al. (in prep.) model. For the sub-Jupiter mass planets, we used the same power-law distribution for the masses, extrapolating it from the super-Jupiter ( >1M_{J} ) regime into the sub-Jupiter ( <1M_{J} ) regime. For the separation distribution, we use a model where we double the super-Jupiter occurrence rate within 10 AU but maintain a flat distribution in \log(a) to the outer cutoff radius (orange curve in Fig. 8 ). This is because previous work by Fulton et al. 2021 showed that the occurrence rate of sub-Jovian planets (0.1 - 1 M J ) is enhanced by almost twice compared to that of the super-Jupiters for orbits between 1 and 10 AU, but unknown beyond 10 AU range. While this survey may not capture this sub 1M_{J} discovery space, large-scale JWST programs, like Carter et al. 2024 , will shed light on this population of sub-Jupiter and sub-Saturn mass exoplanets beyond 10 AU. We derive the companion occurrence rate by integrating over the mass-orbital space over the range of masses defined above and separations from 0 out to an outer cutoff radius constrained by the SPHERE FOV. The estimated total frequencies of sub-Jupiters, super-Jupiters, and brown dwarfs within the FOV are 10.6 \% , 5.6 \% , and 12.4 \% , respectively. These occurrence rates are consistent with estimates from the literature (e.g., Meyer et al. 2018 ; Susemiehl & Meyer 2022 ). For companion generation, we created 10^{5} random sets of orbital parameters using Poisson statistics and the following priors: uniform priors for the longitude of the ascending node and the longitude of periastron, cosine priors for inclination, and Gaussian priors for eccentricity ( \mu = 0, \sigma = 0.3, Hogg et al. 2010 ). For each set of orbital parameters, we simulated 360 positions evenly spaced in time and assessed detectability across an even grid of companion masses and semimajor axes. The same set of orbital parameters was applied to every target. For low-mass planets, we utilized the BEX evolutionary models ( Linder et al. 2019 ), while for higher-mass companions, we used the ATMO 2020 models ( Phillips et al. 2020 ). The results from the modeling are illustrated in Fig. 9 , which shows the simulated populations of brown dwarfs and giant planets around M dwarfs against the SPHERE contrast curves. Based on the parametrisation described above, the code predicts that our HCI observations would detect no Sub-Jupiters, \sim 0.01 Giant Planets and \sim 0.26 Brown Dwarfs in our sample. This is consistent with our non-detections. Figure 8: Frequency distribution for models. The semi-major axis distribution for generated companions follow a lognormal distribution for the giant planet model from Meyer et al. in prep., a lognormal joint with a log-uniform distribution beyond 10 AU for the sub-Jupiter model, and the brown dwarf model uses Winters et al. 2019 , corrected to physical separation. Here, the sub-Jupiter model is a customized model where we have assumed the frequency within 10 AU is double that of the super-Jupiter/gas giant population, but log-flat beyond 10 AU. We normalise the distributions according to Meyer et al. in prep. and integrate over the mass-orbital space to estimate companion frequencies out to a certain radius. Figure 9: Comparison of the depth of search of the SPHERE H2 survey for the presented 7 M dwarf sample (Fig. 6 ) with a population of 100’000 draws from our parametric model for companions around M dwarfs. The contour lines give the numbers of stars around which the survey is sensitive to substellar companions as a function of mass and semimajor axis. The brown dwarf popluation generated from the model is represented with shades of blue (high density) to white (low density), while the giant planet part of the model is represented as a heatmap from red (low density) to yellow (high density of companions). 5 Discussion and outlook In this study, we have demonstrated that combining RV, HCI, and astrometric observations significantly enhances the constraints on the parameter space for potential companions, compared to the results obtained with any individual technique. We have quantified this improvement and found that the percentage of detectable companions within the considered parameter space ( a<100 AU, m {}_{\rm{c}}<50 M J ) increases by up to \sim 60\% compared to RV alone (improvement achieved for GJ 3325), \sim 50\% compared to HCI alone (for GJ 3325), and \sim 12\% compared to astrometry alone (for GJ 367). The relatively lower performance of RV is not surprising, as the parameter space considered in this study, both in terms of extension and resolution, has been optimised for the SPHERE observations, whereas RV techniques are generally more sensitive to companions at smaller semi-major axes. In contrast to Boehle et al. 2019 , we found no significant correlation between the magnitude of improvement and specific target or observational properties, such as stellar distance or the RV time baseline. A similar study focusing on a different region of parameter space, specifically in the planetary regime with shorter orbital periods and lower companion masses, using RV and astrometric data only, is currently ongoing (Sartori, Di Lenia et al., in prep). In the current study we assumed that every point in the parameter space, i.e. every semi-major axis / companion mass pair, is in principle equally likely. Specifically, we did not consider any prior based on current statistics of known detections (e.g., from Bowler 2016 ; Meyer et al. 2018 ; Vigan et al. 2021 ; Bergsten et al. 2022 ). In this way, we do not have inherited biases due to the sample selection and observational design of those studies. We also did not consider any potential information about the systems architecture from previous observations of our targets. For example, Goffo et al. 2023 derived an inclination i_{\rm b}=79.89\pm 0.85\deg based on the TESS transit light curves of GJ 367. For this specific source, fixing the inclination to this high value would increase the percentage of possible detections with RV and astrometry ( +16.6\% and +8.2\% , respectively) and decrease the percentage for HCI ( -13.1\% ). The systematic inclusion of this type of prior information could be considered in the future as a further refinement of our method. The presented method and results rely on already available data, specifically from observations which are (mostly) non-detections. Even in the now dawning ‘‘era of big data’’ 9 9 9 New and upcoming telescopes and surveys such as the Vera Rubin Observatory ( Ivezic et al. 2008 ; LSST Science Collaboration et al. 2009 , formerly known as LSST) and the Square Kilometer Array (SKA, Braun et al. 2019 ) will provide hundreds of PB of data, unprecedented in the history of astronomy. , a re-analysis of archival observations with new methods and post-processing techniques is very valuable to extract important information and statistic about planetary and (sub-)stellar architecture even from data which were previously discarded. When planning new observations to search for companions, however, it is important to consider not only the parameter space explored (as it can be done with our method), but also the accessibility and cost of the data in terms of availability of telescopes and surveys, as well as the required observing time. While the PMa has to be computed based on two astrometric surveys (e.g., Brandt 2018 ; Brandt 2019 ; Kervella et al. 2019a ; Kervella et al. 2019b ; Kervella et al. 2021 ; Kervella et al. 2022 ), in this case Hipparcos and Gaia, and is therefore only available for targets in those catalogues, RV and HCI observations can be obtained for many targets from currently operating telescopes. The sensitivity of RV strongly depends on its time baseline (e.g., Sartori et al. 2023 ), and requires multiple observations over years or even decades in order to confidently constrain the period and (minimum) mass of potential companions orbiting at large distances from the star. On the other hand, the constrast reached by HCI depends, among other factors, on the exposure time and on the amount of FOV rotation 10 10 10 The FOV rotation is crucial for the applicability of post-processing techniques such as ADI, e.g., Marois et al. 2006 (e.g., Follette 2023 ). From the data analysis point of view, while HCI requires advanced post-processing techniques in order to robustly compute the contrast curves and confirm potential detections (e.g., Amara & Quanz 2012 ; Stolker et al. 2019 ; Bonse et al. 2023 ; Bonse et al. 2024 ), RV is more affected by stellar variability, which is often difficult to disentangle from the companion signal (e.g., Meunier 2021 ). While SPHERE is one of the best instruments for HCI to date (e.g., Follette 2023 ), future studies using e.g., the Enhanced Resolution Imager and Spectrograph (ERIS, Davies et al. 2018 ) at the VLT and the James Webb Space Telescope ( JWST ; Gardner et al. 2006 ) would offer significant improvements in both resolution and sensitivity. ERIS, with its advanced near- and mid-IR AO enabled imager NIX ( Parker et al. 2016 ; Boehle et al. 2021 ), will improve the contrast performance at small angular separations. Notably, observations in the L and M bands (3-5 \mu m) will be particularly effective in detecting cooler, lower-mass companions such as giant planets and brown dwarfs. Additionally, the JWST Near-InfraRed Camera (NIRCam, Rieke et al. 2005 ; Rigby et al. 2023 ), which has a similar wavelength coverage and angular resolution as NIX, and the Mid-Infrared Instrument (MIRI, Glasse et al. 2015 ; Wright et al. 2023 ) will offer an unprecedented sensitivity in the near- and mid-IR (0.6–28 \mu m), enabling the detection of even cooler and more distant companions. The combination of ERIS ground-based high-angular-resolution imaging and JWST space-based sensitivity will significantly extend the parameter space for detecting and characterizing substellar companions, especially in cases where SPHERE H-band data might have missed fainter or cooler objects due to contrast limitations at smaller separations or at longer wavelengths. Moreover, combining archival HCI observations with new observations will provide a longer temporal baseline, allowing us to follow the orbital motion of (potential) companions between epochs, and thus to detect companions that were not visible in one of the observations because of their current position relative to the star. This is particularly important for nearby stars, such as those in our sample, where companions exhibit (apparent) larger orbital motions on short timescales, leading to improved detection completeness and better constraints on orbital parameters. The RV measurements used in this study are from archival HIRES and HARPS observations, with time baselines ranging from 4090 days (GJ 1125) to 6961 days (GJ 332). In future studies, newer RV instruments such as the Echelle SPectrograph for Rocky Exoplanets and Stable Spectroscopic Observations (ESPRESSO, Pepe et al. 2021 ; González Hernández et al. 2018 ) at the VLT, and other upcoming high-precision spectrographs, will provide significant improvements in sensitivity compared to these archival data. ESPRESSO, with its ability to achieve an RV precision down to the cm s -1 level, is particularly suited for detecting lower-mass companions, including super-Earths and giant planets in longer-period orbits, which may have been undetectable with previous instruments due to RV jitter and instrumental limitations. In addition, other instruments like EXtreme PREcision Spectrograph (EXPRES, Jurgenson et al. 2016 ) and the Near Infra Red Planet Searcher (NIRPS, Bouchy et al. 2017 ; Wildi et al. 2017 ), which operate in the optical and near-infrared regimes, respectively, will complement ESPRESSO by extending RV coverage to different wavelength ranges, reducing the impact of stellar activity on RV measurements, and improving sensitivity to cooler and low-mass companions around M-dwarfs. The increased precision, broader wavelength coverage, and improved stability offered by these next-generation RV instruments will allow us to probe a wider range of parameter space for detecting substellar companions, particularly for nearby stars, where the RV signal is stronger due to proximity. The combination of archival data with new high-precision RV measurements will also allow us to constrain long-term trends more robustly and identify companions with longer orbital periods that may not have been detectable in previous RV datasets. Combining observations obtained with different techniques is crucial for population statistics, as it allows for the exploration of a wider parameter space and helps mitigate observational biases inherent to each method. This multi-technique approach leads to a more complete understanding of stellar-companion architectures and demographics across a broad range of parameters, such as semi-major axis, companion mass, and host star spectral type. In this study, we focus on the combination of RV, HCI, and astrometry, building on the work of Boehle et al. 2019 , who analyzed a similar sample but considered only RV and HCI. Nearby stars are however also prime targets for large monitoring surveys such as Kepler ( Borucki et al. 2010 ) and the Transiting Exoplanet Survey Satellite ( TESS ; Ricker et al. 2015 ), which have the sensitivity to detect exoplanet transits and transit timing variations (TTV, e.g., Agol et al. 2005 ) for thousand of stars on the whole sky. These surveys complement RV and HCI data by probing shorter-period companions, especially for bright, nearby stars. Additionally, microlensing surveys provide valuable constraints on intermediate-separation and low-mass planets around low-mass stars, primarily M-dwarfs (e.g., Gould et al. 2010 ; Meyer et al. 2018 ). Incorporating data from such complementary techniques will further expand the parameter space covered by our combined approach, yielding a more robust understanding of planetary system architectures. As we elaborated in Section 4.3 , the fact that we did not detect any new companion is not surprising given the current statistic of known large planets and brown dwarfs (e.g., Bowler 2016 ; Meyer et al. 2018 ; Vigan et al. 2021 ), which shows that large planets on wide orbits, as probed by this study, are quite rare. Indeed, based on such statistics, <1 companion is expected for our sample. These results are also consistent with the parametric model, which predicts no Sub-Jupiters, \sim 0.01 Giant Planets and \sim 0.26 Brown Dwarfs in our sample in the parameter space probed by the HCI data. As we mentioned above, as it includes only 7 targets, our sample cannot be considered a statistical sample yet. As part of the PSION project we are however working on increasing the number of sources for which our method can be applied. Once the sample is big enough, it will be interesting to compare the results, in particular the depth of search and the 2D sensitivity for the whole sample, to population synthesis models based on both gravitational instability (e.g., Boss 1998 ; Vorobyov 2013 ) and core accretion (e.g., Goldreich & Ward 1973 ; Pollack et al. 1996 ) formation mechanisms. As it has been done in the past for different samples (e.g., Janson et al. 2011 ; Mordasini et al. 2017 ; Vigan et al. 2021 ), comparing such models to observations will help discriminating and better understanding the formation process of giant planets and brown dwarfs. As our method allows to directly compare the parameter space probed by the different detection techniques, this comparison will also help understanging on which technique we should concentrate for future observations in order to better constrain such formation mechanisms In summary, this study demonstrates the effectiveness of multi-method approaches in constraining companion populations, and highlights the still unexploited potential of archival data that may have previously been overlooked due to non-detection. Although the small sample size limits the statistical significance of the results presented in this work, the methodology can be easily extended to larger samples and different regions of the parameter space. In this study, we focus on giant planets and brown dwarfs, for which constraining the parameter space is essential for understanding their formation and evolution. The identification of outer giants (like Jupiter in the Solar System) may be essential for the habitability of inner terrestrial planets, e.g. through early water supply by scattering of volatile-rich solids, and later as a safeguard against bombardment (e.g. Sánchez et al. 2018 , Zain et al. 2018 ). Proving the absence of giants near the HZ is also important, as giants near the HZ would make orbits inside the HZ unstable. In the future, applying a similar approach in the planetary regime, particularly for rocky exoplanets, will allow us to better constrain the parameter space for Earth analogues and other terrestrial planets. This in turn will inform the target lists for future space missions such as the Large Interferometer for Exoplanets (LIFE) and the Habitable Worlds Observatory (HWO), which aim to identify potentially habitable or even inhabited planets around nearby stars Acknowledgements. We thank the anonymous referee whose comments helped to improve the quality and readability of this paper. Based on observations collected at the European Southern Observatory under ESO programme 0104.C - 0336. This work has been carried out within the framework of the National Centre of Competence in Research PlanetS supported by the Swiss National Science Foundation under grants 51NF40 \_ 182901 and 51NF40 \_ 205606. The authors acknowledge the financial support of the SNSF. L.F.S. thanks Gabriele Cugno for helpful discussions on high-contrast imaging. Authors contributions. L.F.S.: Conceptualization, Methodology, Software, Validation, Formal Analysis, Investigation, Data Curation, Writing – Original Draft, Visualisation; S.P.Q.: Conceptualisation, Resources, Writing – Review and Editing, Funding Acquisition; M.J.B.: Software ( applefy ), Writing – Original Draft, Review and Editing; Y.L.: Formal Analysis (parametric model), Software, Writing – Original Draft, Review and Editing; F.A.D.: Writing – Review and Editing; C.L.: Conceptualisation, Writing – Review and Editing; A.B.: Conceptualisation, Methodology, Data Acquisition. Software. This research made use of: Astropy 11 11 11 https://www.astropy.org/ , a community-developed core Python package for Astronomy ( Astropy Collaboration et al. 2013 ; Astropy Collaboration et al. 2018 ); Matplotlib 12 12 12 https://matplotlib.org/ ( Hunter 2007 ); PynPoint 13 13 13 https://pynpoint.readthedocs.io/en/latest/ ( Amara & Quanz 2012 , Stolker et al. 2019 ); species ( Stolker et al. 2020 ); applefy 14 14 14 https://applefy.readthedocs.io/en/latest/ ( Bonse et al. 2023 ). Appendix A Initial data Figure 10: Residuals images computed with PynPoint ( Amara & Quanz 2012 ; Stolker et al. 2019 ) for the SPHERE/ H2 observations using full-frame PCA and 100 components. The circular feature at \sim 1 arcsec corresponds to the limit of the AO correction (0.8 arcsec, Fusco et al. 2006 ; Fusco et al. 2016 ). The white bars correspond to 5 AU at the distance of the star. No companion is detected. Appendix B Combined statistic plots Figure 11: Same constraints on companion mass and semi-major axis as shown in 4 but for GJ367. Figure 12: Same constraints on companion mass and semi-major axis as shown in Fig. 4 but for GJ382. Figure 13: Same constraints on companion mass and semi-major axis as shown in Fig. 4 but for GJ3325. Figure 14: Same constraints on companion mass and semi-major axis as shown in Fig. 4 but for GJ402. Figure 15: Same constraints on companion mass and semi-major axis as shown in Fig. 4 but for GJ465. Figure 16: Same constraints on companion mass and semi-major axis as shown in Fig. 4 but for GJ357.