Direct Imaging Discovery and Dynamical Mass of a Substellar Companion Orbiting an Accelerating Hyades Sun-like Star with SCExAO/CHARIS

We present the direct-imaging discovery of a substellar companion in orbit around a Sun-like star member of the Hyades open cluster. So far, no other substellar companions have been unambiguously confirmed via direct imaging around main-sequence stars in Hyades. The star HIP 21152 is an accelerating star as identified by the astrometry from the Gaia and Hipparcos satellites. We have detected the companion, HIP 21152 B, in multi-epoch using the high-contrast imaging from SCExAO/CHARIS and Keck/NIRC2. We have also obtained the stellar radial-velocity data from the Okayama 188cm telescope. The CHARIS spectroscopy reveals that HIP 21152 B's spectrum is consistent with the L/T transition, best fit by an early T dwarf. Our orbit modeling determines the semi-major axis and the dynamical mass of HIP 21152 B to be 17.5$^{+7.2}_{-3.8}$ au and 27.8$^{+8.4}_{-5.4}$ $M_{\rm{Jup}}$, respectively. The mass ratio of HIP 21152 B relative to its host is $\approx$2\%, near the planet/brown dwarf boundary suggested from recent surveys. Mass estimates inferred from luminosity evolution models are slightly higher (33--42 $M_{\rm{Jup}}$). With a dynamical mass and a well-constrained age due to the system's Hyades membership, HIP 21152 B will become a critical benchmark in understanding the formation, evolution, and atmosphere of a substellar object as a function of mass and age. Our discovery is yet another key proof-of-concept for using precision astrometry to select direct imaging targets.

Recent work shows the advantage of targeting stars that show evidence for the dynamical pull of a substellar companion, which provides a complementary approach to blind surveys. For example, targeted high-contrast imaging observations of the nearby Sun-like star HD 33632A from the Hipparcos-Gaia Catalogue of Accelerations (HGCA; Brandt 2018Brandt , 2021 have revealed a brown dwarf (BD) companion in the system (Currie et al. 2020a). The HGCA lists all nearby stars with significant proper motion (PM) accelerations and allows to select promising targets for high-contrast imaging, since the accelerated PM of a star can be caused by its companion. In addition, the HGCA is useful for analyzing the orbits of companions by combining it with DI data and/or radial velocity (RV) measurements, often leading to a ∼10% dynamical constraint on the companion's mass (e.g., Currie et al. 2020a;Bowler et al. 2021;Brandt et al. 2021). Thus, the use of HGCA also enables placing constraints on stellar and substellar evolution models by comparing the model-based mass of a companion with its dynamical mass measurement.
Imaged substellar companions around accelerating stars become even better benchmark objects if key system properties such as age and metallicity are well determined. The Hyades open cluster is one of the most extensively examined open clusters (OCs) in all of astronomy, with a thoroughly vetted membership list, well constrained age, and well determined metallicity (e.g., Brandt & Huang 2015;Gagné et al. 2018;Gossage et al. 2018). With typical distances of about 50 pc, Hyades members are near and bright enough that HGCA is well suited for identifying substellar companions.
We report the discovery of an L/T-transition BD companion around the accelerating star HIP 21152 1 , with the companion's dynamical mass estimation. It is the first substellar companion directly imaged around a Sunlike star in the well-characterized Hyades OC and represents a new benchmark to better understand the properties of substellar objects.  (Hoffleit 1964) with an estimated mass of ∼ 1.3 M (David & Hillenbrand 2015). For this star, we first found a substantial deviation from simple linear kinematic motion (i.e., acceleration) from the HGCA based on Gaia DR2 (Brandt 2018). The updated measurement of acceleration in the HGCA based on Gaia EDR3 (Brandt 2021) is calculated to be χ 2 = 174.6, consistent with a 13.0-σ significance with 2 degrees of freedom (2 DOF). The Banyan-Σ (Gagné et al. 2018) algorithm 2 provides HIP 21152 an extremely high membership probability (99.5%) for Hyades with the inputs from the Gaia EDR3 catalogue 1 The discovery of HIP 21152 B was independently reported by Bonavita et al. (2022) with their VLT/SPHERE imaging performed among a survey for a large sample of accelerating targets. Franson et al. (in prep) have also independently discovered HIP 21152 B with their originally obtained data, and will characterize this system in detail with all the available data. 2 http://www.exoplanetes.umontreal.ca/banyan/ Note-θv represents the characteristic seeing measurements from the Canada France Hawaii Telescope seeing monitor. The integration time of each exposure, the numbers of exposures used in our analysis, the total variation of parallactic angle in each sequence are represented by texp, Nexp, and ∆P ar, respectively. The column of "Data Proc." describes the types of our ALOCI PSF subtractions. S/N represents the companion PSFs' signal-to-noise ratios calculated from the 22-channel collapsed images.

SCExAO/CHARIS and Keck/NIRC2
High-Contrast Imaging We performed high-contrast imaging observations using adaptive optics (AO) on the Subaru and Keck II telescopes between October 2020 and October 2021, in photometric and good-to-average seeing nights (Table  1). The Subaru observations utilized AO188 (Hayano et al. 2008) for first stage correction of atmospheric turbulence, followed by a faster higher order correction of residual wavefront errors by the extreme AO system SCExAO (Jovanovic et al. 2015;Currie et al. 2020b). A coronagraph within SCExAO is then deployed to mask the central starlight, yielding high contrast images that are captured by the CHARIS integral field spectrograph (IFS; Groff et al. 2016). In our Keck observations, the target lights corrected by a near-IR Pyramid wavefront sensor were transferred to the NIRC2 camera (Bond et al. 2020).
The SCExAO/CHARIS and Keck/NIRC2 data obtained in 2020 have 55-98 minutes of on-source integration time. Our shallower SCExAO/CHARIS data set obtained in October 2021 aimed solely at rejecting the possibility that the companion candidate is a background object. All observations were performed in angular differential imaging (ADI) mode, and CHARIS's IFS also enabled spectral differential imaging (SDI; see Op-penheimer & Hinkley 2009, and references therein). All CHARIS data were taken with a low-resolution (R ∼ 18) spectroscopic mode to obtain wide wavelength coverage and the Lyot coronagraph with a 0. 23 diameter mask; NIRC2 data were taken in the L -band filter also using a Lyot coronagraph but with a larger (0. 6) mask (see Table 1). By modulating SCExAO's deformable mirror, we generated four satellite spots around the point spread function (PSF) of HIP 21152 to enable astrometric and spectrophotometric calibration (Sahoo et al. 2020), while the NIRC2 coronagraph allows a direct stellar centroid estimate due to its partial transparency. To flux-calibrate the NIRC2 data, we obtained unsaturated PSFs of the star before and after the coronagraphic sequence.
Using the pipeline of Brandt et al. (2017), the raw CHARIS data were calibrated, and converted into 2Dimage cubes consisting of 22 wavelength channels. To further process these extracted data cubes, we used the CHARIS Data Processing Pipeline 4 following the outline in Currie et al. (2020b); the details of our highcontrast image processing are provided in Appendix A. The spectra of HIP 21152A were measured with the satellite spots in each channel for spectrophotometric calibration, where we adopted an F5V model atmosphere from the Kurucz library (Castelli & Kurucz 2003) with the star's 2MASS photometry (Skrutskie et al. 2006(Skrutskie et al. , 2019. The satellite spots are also used to register the central star's PSFs to a common center. While the four spots have roughly equal brightnesses in each wavelength slice, the spot intensities for the December 2020 data showed a large systematic variation, making spectrophotometric calibration for these data more uncertain (see Appendix B.1). We processed the NIRC2 images using a well-tested pipeline (Currie et al. 2011(Currie et al. , 2014) that carries out standard steps of sky subtraction, image registration, photometric calibration, and PSF subtraction. Our image processing primarily adopts the ALOCI algorithm for ADI PSF subtractions (e.g., Currie et al. 2014Currie et al. , 2018. We attempted additional processing to improve or validate the fiducial reductions using alternate ADI reductions that adopt proprietary version of ALOCI and SDI speckle suppression (see also Appendix A). Figure 1 shows the HIP 21152 images obtained from our four data sets, from which HIP 21152 B is detected at signal-to-noise ratios 5 (SNRs) of 10-19. We achieve comparable detections with the proprietary version of ALOCI, although the throughput of the proprietary version is far higher. SDI increases the SNR of the detection at the expense of greater spectroscopic uncertainty. To correct our spectrophotometry and astrometry for biasing due to processing, we carried out forward modeling as in previous work (Currie et al. 2018). The forward modeling on the ADI+SDI (ASDI) PSF subtraction accounts for the companion's spectral type.

High-Resolution Doppler Spectroscopy
We monitored HIP 21152A with the high-efficiency fiber-link mode of the HIDES spectrograph equipped on the Okayama 188cm telescope (HIDES-F; Kambe et al. 2013) to measure the star's radial velocities (RVs). Our monitoring was conducted for about one year from 2011-12-30 and two years from 2020-02-11. In December 2018, the spectrograph was re-arranged to improve the stability of RV measurements against temperature fluctuations. We used an image slicer, setting the spectral resolving power to be 55,000 by a 3.8-pixel sampling. The spectra of HIP 21152 passed through an I 2 cell, whose absorption features imposed on the spectra are used as references for RV calibration. Except for three poor-SNR (< 30) spectra, we obtained 32 I 2 -imposed spectra of HIP 21152A with integration times (IT) of 900 or 1800 seconds, and four I 2 -free spectra at various nights (IT = 1800 × 4 seconds). The I 2 -imposed spectra have SNRs ranging from 57 to 258 at ≈5500Å. Our RV calculations adopt a wavelength range from 5028 to 5753 A, which contains numerous I 2 features and little telluric absorption. The data calibrations and extractions 5 The SNRs were calculated with the correction of Mawet et al. (2014).
of one-dimensional spectra were performed in a standard way based on IRAF. The one-dimensional I 2 -free spectra are combined into a single template spectrum after removing outliers, applying 3-pixel median smoothing 6 and barycentric correction to each spectrum. The same master template spectrum was compared with each of the 32 I 2 -imposed spectra to measure the RVs of HIP 21152 without producing an offset in the measurements. Our RV measurements were obtained using the pipeline of Sato et al. (2002Sato et al. ( , 2012, which corrects the line profile fluctuations originating from the instrumental instability by modeling them from I 2 absorption lines. The spectra were divided into several segments with wavelength widths of ≈5.3-6.1Å and the RVs were calculated in each segment. The wavelength widths of each segment were set to be much wider than standard RV measurements in HIDES because of HIP 21152's rapid rotation. The adopted widths provide the smallest RV errors among several attempted segment widths. The segment-by-segment RVs were statistically summarized to be the final RV measurements in Appendix C.

INFRARED COLORS, SPECTRUM, AND ATMOSPHERE OF HIP 21152 B
We base the following discussions on the 2020-October spectra (after correcting for spectrophotometry bias) presented in Appendix B.3 because the data at this epoch have the highest SNR, the most stable PSF quality, and the best calibration (see Appendix B.1). We calculated J-, H-, and K s photometry from the 2020-October ADI spectrum using the bandpasses' filter transmission profiles: J = 17.72 ± 0.20, H = 17.04 ± 0.15, and K s = 16.55 ± 0.17 mag. From NIRC2 data, we measure L = 15.04 ± 0.12 for HIP 21152 B 7 . Figure 2 compares the near-infrared colors of HIP 21152 B to those of directly imaged BDs and young exoplanets. HIP 21152 B's colors are best reproduced by early T dwarfs, near the L-T transition, and are slightly bluer than HD 33632Ab.
The extracted spectra of HIP 21152 B are shown in Figure 3, where we plot both the ADI and ASDI spectra as well as the ADI spectrum reduced with a proprietary code; measurements in each spectral channel agree between all reductions. HIP 21152 B's spectral shape shows strong absorption attributed to water opacity at the gaps between major near-infrared filters and a very blue slope at 2.2-2.4 µm consistent with methane absorption.
To more quantitatively determine HIP 21152 B's spectral type, we performed a least-square analysis by comparing the spectrum of HIP 21152 B with the template spectra of cool dwarfs (Currie et al. 2018). The template spectra were taken from the Montreal Spectral Library 8 (e.g., Gagné et al. 2015;Robert et al. 2016). We analyzed the 2020-October spectrum of HIP 21152 B obtained with the ADI-based PSF subtraction. The fit accounted for the spatially and spectrally correlated noise in an IFS spectrum using the scheme developed by Greco & Brandt (2016). We found that 95% off-diagonal elements of the spectral covariance were smaller than ≈0.16 at the companion's angular separation, indicating that the noise is only weakly correlated both spatially and spectrally. The six wavelength channels affected by significant telluric absorption were omitted in the fit. The χ 2 -based comparison shown in Figure 3 compares HIP 21152 B's spectrum to selected objects from the Montreal library. Overall, it is best fit by the T1.5dwarf SIMP J2215+2210. Furthermore, the earlier-type template spectrum that best matches HIP 21152 B is the T0.5 object SIMP J1200-2836 (∆χ 2 2), while the later-type best match is T2 (∆χ 2 2; SIMPJ 1629+0335). Alternate ADI reductions find similar results. Earlier L-type dwarfs predict troughs in the water bands bracketing J, H, and K to be too shallow and have slopes in the 2.2-2.4 µm range too red to be consistent with HIP 21152 B. Hence, we estimate the most-likely spectral type of the companion as T1.5 +0.5 −1.0 , which is slightly later-type than HD 33632Ab (L9.5 +1.0 The three lines correspond to the fiducial ADI, ASDI, and alternate ADI (labeled as "alt") reductions. One-σ error bars are appended to the ADI spectra. (Right) Comparisons between the fiducial spectrum of HIP 21152 B (blue dashed lines) and three template spectra from the Montreal Spectral Library (gray solid lines). The wavelengths where telluric absorption is significant are masked and were not used in our spectral typing. The reduced chi-square values (χ 2 ν ) computed by least-square fitting (15 DOF = sixteen data points minus one optimized parameter for spectrum scaling) are shown above each comparison with the names of the compared BDs and their spectral types.

HIP 21152 B Astrometry
We measured the projected separations (ρ in unit of milli-arcseconds or mas) and the position angles (PAs) of HIP 21152 B relative to its central star by fitting elliptical Gaussian models to the companion PSFs identified in all the ALOCI-processed images. For the CHARIS images, the PSF-fitting was conducted in the images after median-combining all the wavelength channels. Forward-modeling allowed us to assess astrometric biasing due to processing. Table 1 shows the HIP 21152 B astrometry. The astrometric errors were estimated by taking into account contributions from noise including speckle residuals, calibration errors on the plate scales and true-north angles of our used instruments, and systematic errors in the measurements of the primary star's absolute centers (see Appendix D.1 for detail and the error budgets).
HIP 21152 B's motion is inconsistent with the relative motion of a background object expected from the astrometry information of the central star from Gaia EDR3 (Gaia Collaboration et al. 2021): the measured vs. predicted position in October 2021 differs by more than 10σ in both ρ and PA (see also Appendix D.2).

Orbit and Dynamical Mass Estimates
To constrain HIP 21152 B's orbit and dynamical mass, we model the star's absolute astrometry from HGCA, the star's RV measurements, and the companion's relative astrometry from direct imaging, using the orvara software (Brandt et al. 2021). Our orvara analysis carries out Markov Chain Monte Carlo (MCMC) simulations adopting 15 temperatures in the parallel tempering chain and 100 walkers. Each chain has 7×10 5 steps; we save every 25th step for 28,000 steps per walker. We assume a Gaussian prior for the stellar mass: its mean and standard deviation are set to be 1.3 and 0.1 M following David & Hillenbrand (2015). For the other parameters, our fit assumed the default priors of orvara including 1/M prior for the companion's mass (Brandt et al. 2021, see also Appendix D.2). RV jitter was simulated in the range of 0-100 m s −1 with a log-flat prior. We discarded the initial 2,500 steps as burn-in phase from the recorded 28,000 steps. With 100 walkers, we have 2.55 × 10 6 samples for inference. Figure 4 shows the corner plot of fitted system parameters from orvara, predicted orbits, and predicted RVs (see Appendix D.2 for the fits to the other measurements). The median and the 16-84th percentiles of the MCMC posteriors are provided in Appendix D.2. HIP 21152 B has a best-fit semi-major axis (a) of 17.5 +7.2 −3.8 au, viewed at a high inclination of i = 104.8 +15 −6.9 • . The estimated mass of the primary largely reflects our input prior, while the companion's mass was estimated to be 27.8 +8.4 −5.4 M Jup . We find no strong constraints on the eccentricity (e) of the companion posterior.

DISCUSSION
We directly imaged a substellar companion orbiting ∼18 au from the Sun-like star HIP 21152, which has an accelerating proper motion. The companion's spectrum is best reproduced by an object near the L/T transition, plausibly an early T dwarf. The system is a member of the Hyades open cluster (OC) which has a wellconstrained age of 750 ± 100 Myr. It is notable that there have been the reports of single BDs (e.g., Lodieu et al. 2019) and BD binaries (e.g., Duchêne et al. 2013) directly imaged in this OC. In contrast, there has been no unequivocal confirmation of directly-imaged companions that are less massive than the hydrogen burning limit (e.g., Fernandes et al. 2019), around main-sequence stars in Hyades (see a note in Appendix E). Accordingly, HIP 21152 B is a crucial benchmark to understand substellar-mass objects as below.
HIP 21152 B's dynamical mass is approximately twice the deuterium burning limit and at/slightly above the estimated turnover in mass separating massive jovian exoplanets from BDs (Sahlmann et al. 2011). Table 1 in Franson et al. (2022) lists all the ages and dynamical masses of directly-imaged substellar companions. We can compare those companions with HIP 21152 B, which has a fractional age uncertainty of 13% (100 Myr / 750 Myr). Smaller fractional uncertainties in age estimations lead to smaller fractional uncertainties in mass estimations for directly-imaged substellar companions. For example, the models of Baraffe et al. (2003) convert a luminosity of log (L/L ) ≈ −4.9 to 8 ± 1 M Jup (26 ± 4 M Jup ) at an age of 50 ± 10 Myr (500 ± 100 Myr); meanwhile, the same luminosity corresponds to 8 ± 2 M Jup (26 +6 −7 M Jup ) at 50 ± 20 Myr (500 ± 200 Myr). The corresponding fractional uncertainties for all cases in the list of Franson et al. (2022) are larger than 16%, except for the planets around β Pic, which are 13% just as for HIP 21152 B. Furthermore, the majority (11/18) of the companions have fractional uncertainties higher than ∼30%. Thus, the smallest fractional age uncertainty of HIP 21152 B provides the highest fidelity model-dependent mass estimations, besides β Pic bc.
In the list, β Pic bc and HR 8799 e are the only directly imaged giant planets whose masses have been dynamically constrained. HIP 21152 B is the closest to those benchmark planets in terms of dynamical mass, best helping unveil the physical and chemical connection between giant planets and BDs. We note that only HIP 21152 B is firmly associated with an OC among these benchmark substellar companions. The methods used to characterize stellar and substellar objects such as age estimation techniques (e.g., Mamajek & Hillenbrand 2008) and evolutionary models (e.g., Tognelli et al. 2020) have been calibrated by the observations of OCs. The Hyades OC is especially useful for such calibrations due to its proximity to the Sun. Hence, HIP 21152 B would be available as one of the best benchmark companions to test evolution and atmosphere theories of cool objects among directly-imaged BDs with inferred dynamical masses.
In contrast to the dynamical mass, given HIP 21152 B's bolometric luminosity of log (L/L ) = −4.673 ± 0.066 and a system age of ≈ 650-850 Myr, the Baraffe et al. (2003) luminosity evolution models yield slightly higher predicted masses of 33-42 M Jup . However, temperatures implied by this age range -1200-1350 K -are broadly consistent with spectroscopically-derived values. Thus, HIP 21152 B may provide another example of 1-2σ tension between substellar dynamical masses and those inferred from luminosity evolution at intermediate ages. Dupuy et al. (2009Dupuy et al. ( , 2014 found significant discrepancies between the dynamical masses and luminosity evolution-inferred masses for the BD binaries GJ 417 BC and HD 130948 BC. Both systems have ages comparable to Hyades members but have masses of ≈ 50 M Jup , or ∼ 50% higher than HIP 21152 B. Future astrometric monitoring of HIP 21152 B will further tests that luminosity evolution models can overestimate the masses of substellar objects. HIP 21152 B is a benchmark to test atmosphere models of substellar objects as well. For instance, gravitysensitive absorption features such as K I and FeH can be measured via medium-to-high resolution spectroscopy (e.g., Martin et al. 2017). Those measurements allow the comparison of HIP 21152 B's surface gravity estimated by atmosphere models with that constrained by its dynamical mass and radius (which is appropriately assumed to be about 0.1 R at Hyades's age; Baraffe et al. 2003). It is also interesting to characterize HIP 21152 B in the context of L/T transition of substellar objects depending on surface gravity and metallicity (e.g., Faherty et al. 2012   characterization as an anchor point, due to a super-solar metallicity expected from its membership to the Hyades OC (Gagné et al. 2018;Gossage et al. 2018) and the semi-empirically constrained surface gravity.
HIP 21152 B's companion-to-primary mass ratio, q, is ∼ 2.0% +0.7% −0.4% . This value is intermediate between bona fide directly imaged exoplanets like HR 8799 bcde (e.g., Marois et al. 2008;Currie et al. 2014) and BD companions imaged around Sun-like stars such as HD 33632 Ab (Currie et al. 2020a) and HD 47127 B (Bowler et al. 2021). Very few binary star companions have mass ratios this low (Kraus et al. 2008); surveys suggest that the substellar mass function turns over at a mass ratio of q ∼ 0.025, where lower (higher) mass ratio companions may be best interpreted as exoplanets (BDs).
In OCs, the gravitational interactions of passing stars can perturb companions on wide orbits and cause ejections in some cases. Fujii & Hori (2019) explored the ejection of planets by modeling stellar encounters in OCs via N -body simulations. They found that ejections do not frequently occur in a low-density OC like the Hyades, even in cases where planets orbit their hosts at semi-major axes of 10-100 au. Their findings should be applicable also to low mass-ratio companions like HIP 21152 B and consistent with this discovery, contributing to verifying such a theory for dynamics of planet/BD companions in OCs.
Finally, this discovery provides further evidence of the promise of using precision astrometry to select direct imaging targets. Even for a system 750 ± 100 Myr old, we were able to directly detect a ≈20-30 M Jup companion orbiting on solar system scales with high SNRs, demonstrating the capability of extreme AO instruments to detect cooler companions at 10-20 M Jup on the same scale. A large sample of directly imaged exoplanets and BDs with high quality spectra, dynamical masses, and well-constrained ages will clarify how atmospheres of substellar objects evolve depending on companion mass and how they link to their formation mechanisms. We are honored and grateful for the opportunity of observing the Universe from Maunakea, which has the cultural, historical, and natural significance in Hawaii. We appreciate the critical support from all the current and recent Subaru and Keck Observatory staffs. Their support was essential in achieving this discovery, especially amidst the many difficulties associated with the COVID-19 pandemic.
We  We here describe the specific considerations that were taken during our data reductions for high-contrast imaging. Poor-quality data cubes needed to be removed, since they affect the data reduction procedure. In order to remove poor Strehl-ratio CHARIS data cubes, we processed only the data with peak-to-halo ratios greater than 10, which are the signal ratios of satellite spot peaks relative to halos of the central star PSFs. This criterion led us to exclude 8 (3%) and 69 (35%) data cubes from observations obtained in October and December of 2020, respectively. The October 2021 data are split into two sequences, between which the wind-driven halo changed direction: cubes from one sequence are poorly correlated with the other. For these data, we retain the first sequence since the AO performance and the change in parallactic angle are better (35 data cubes removed).
For all data sets, we spatially filtered the data using a radial-profile subtraction and subtracted the speckle halo using the ALOCI ADI algorithm (Currie et al. 2014(Currie et al. , 2018. For the CHARIS data of October and December 2020, we truncated the set of reference images for each target image based on the correlation between every target-reference pair, selecting the 120 and 100 best-correlated reference images, respectively. We did not apply this truncation to the processing for the data of October 2021 due to the small number of available exposures. Other algorithm parameters defining the geometry over which we optimized our reference PSF construction and subtracted this PSF were varied but were generally close to pipeline default values: an optimization area (in units of PSF footprints) of N A = 100 and a rotation gap of 0.5-0.75 full-width-half-maximum of PSF (see e.g., Pueyo et al. 2012).
To further explore speckle suppression, we considered two additional approaches. First, for CHARIS data we also applied an SDI reduction on the post-ADI residuals as performed in Currie et al. (2018), which improves speckle suppression but may introduce less reliable spectral extractions (Pueyo et al. 2012). Due to the worse observing conditions at October 2021, our analysis relies on SDI reductions for this epoch. Second, we made alternate ADI reductions that are different from the main procedures, using a proprietary version of ALOCI for which modifications included varying the optimization/subtraction zone geometries and turning on/off a pixel mask over the subtraction zone. With the adoption of pixel masking, linear-combination coefficients for reference PSF construction are calculated after masking the pixels in PSF subtraction zones. This technique has been used elsewhere (e.g., Pueyo et al. 2012;Currie et al. 2018) including the public ALOCI pipeline to suppress a bias from companion PSFs and significant self-subtractions; indeed, we obtained better throughputs using this technique. Meanwhile, the proprietary version of ALOCI has the options to adopt several types of optimization zone shape including the standard shape that has been commonly used (e.g., Pueyo et al. 2012) and optimize the zone geometry.

B.1. Spectroscopy from December 2020 Data
Inspection of the December 2020 data showed issues with spectrophotometric calibration that impeded our ability to extract a spectrum with a quality comparable to that obtained from October 2020 data. Using our default ALOCI-ADI reduction, we measure broadband photometry of J = 18.29 ± 0.42, H = 17.15 ± 0.19, and K s = 16.16 ± 0.16 mag. While the H-band photometry at October and December 2020 is consistent within 1σ, the J-and K s -band measurements are discrepant at 1.2σ and 1.7σ levels. As shown in Figure 5 (left panel), the differences in spectra between the October 2020 and December 2020 ADI reductions are significantly larger than error bars in the shortest and (especially) longest wavelength channels.
Further investigation of this issue identified some partial mitigation measures. The alternate ADI/ALOCI reduction using pixel masking and a different optimization zone geometry yields better agreement ( Figure 5 left panel), suggesting that contamination from residual speckles may be affecting the K band measurements.
We also note that the measured spectrum from the December 2020 data can be affected by the ununiform brightnesses of the satellite spots adopted in the spectrophotometric calibration. The satellite spots in the December 2020 cubes showed modest brightness differences at a channel (2.37 µm), while they show smaller brightness differences at 1.58 µm ( Figure 6). The December 2020 satellite spots show the spot-to-spot ununiformity larger (∼10%) than the October 2020 spots only in the 2.37 µm channels ( Figure 6) and telluric-dominated channels, whereas the spots should have roughly equal brightness yielding spectrophotometric precision on the order of ∼2% (Currie et al. 2020b).
Given the apparent problems with the December 2020 data and the higher quality of the October 2020 data, we adopt the October 2020 spectrum as the basis for our analysis 9 . The lower PSF qualities of December 2020 images are also implied by the larger number of data that we needed to omit (see Section A) and the more unstable systematic variations of satellite-spot brightness (see Figure 6). We note that the J − K s colors for December 2020 spectra derived from the alternate ADI reduction are J − K s = 1.60 ± 0.35 with an apparent J-band magnitude of 18.05 ± 0.29 (absolute J band magnitude of 14.87 ± 0.29). The resulting color-magnitude diagram positions also lie at the L/T transition as was found for our analyses of the October 2020 spectrum. Thus, any uncertainties in the shape of HIP 21152 B's spectrum at red wavelengths have a negligible impact on our broad conclusions about HIP 21152 B as a substellar object at the L/T transition.

B.2. Spectroscopy from October 2021 Data
The right panels of Figure 5 compares the spectrum extracted from our October 2021 data set to that taken a year prior. Overall, the 2021 spectrum is noisier at all wavelengths but agrees in both absolute flux density and shape with the 2020-October results. New spectra extracted from higher signal-to-noise data and taken at higher resolution are required to advance our understanding of HIP 21152 B's spectral properties.

B.3. HIP 21152 B Spectrum
In Table 2 below, we present the HIP 21152 B spectrum derived from our standard ALOCI-ADI reduction of the October 2020 data.

C. RADIAL VELOCITY MEASUREMENTS OF HIP 21152
In Table 4, we list individual relative radial-velocity measurements for HIP 21152. As CHARIS was craned in and out of position multiple times between our observations, we reassessed the plate scale and true-north orientation angle offset of the CHARIS detector. As in Currie et al. (2018Currie et al. ( , 2020a, we used Figure 5. (Left) Comparison between October 2020 spectra and December 2020 spectra extracted using different reduction approaches. (Right) October 2021 spectrum compared to spectra extracted from October 2020 data. On both the left and right panels, "alt" indicates the alternative ALOCI-ADI reductions. The spectral wavelengths of the spectra on the left and right panels are slightly shifted to avoid plot overlap.
SCExAO/CHARIS and Keck-II/NIRC2 data of the companion around HD 1160 to identify any change in detector astrometric properties 10 . As described in Currie et al. (2022), these analyses favor a slightly revised plate scale of 16.15 ± 0.05 mas pixel −1 but otherwise no measureable changes in the astrometric calibration determined in Currie et al. (2018). Our analysis is based on the revised pixel scale described above, and the true-north orientation offset angle in Currie et al. (2018).
We also consider an absolute astrometric error due to uncertainties in determining the star's positions. For the Keck/NIRC2 data, Konopacky et al. (2016) quote an uncertainty of ∼ 2 mas in the star's center when determined through the partially transmissive Lyot coronagraph. For CHARIS, internal source tests described in Currie et al. (2020b) included analyses of the star position determined from fitting the satellite spots vs. unobstructed PSF centroids. The tests reveal up to a 0.25 pixel offset (∼4 mas) in reported position of the centroid estimated from satellite spots and that determined from an unobstructed PSF. The NKT Photonics SuperK laser we used for this analysis only extends to 1.7 microns, so we do not have a direct measurement of any biasing at longer wavelengths. The source of this difference is unclear but could be due to residual field distortion. As an empirical test, we compared the centroid positions for HD 1160's companion in 7 different data sets from August 2020 to January 2022, a timeframe  Figure 6. Normalized peak count of satellite spots in the reddest wavelength (2.37 µm) and 1.58 µm slices for cubes in the October 2020 (left) and December 2020 data (middle). The righthand panels label the satellite spots.
over which we expect the orbital motion to be negligible; the standard deviation in the east and north positions are ∼3.7 mas and ∼2.3 mas. To be conservative, we adopt an absolute astrometric error in each coordinate of 0.25 pixels (= 4 mas) for CHARIS. For CHARIS, uncertainties from astrometric biasing due to processing and (for the October and December 2020 data sets) the intrinsic detection SNR are small compared to intrinsic uncertainties in the pixel scale (0.05 mas), north position angle offset (0.27 • ), and absolute centroid measurement (∼ 4 mas). However, residual and only partiallywhitened speckle noise may contaminate centroid measurements more than expected from an SNR estimate (e.g. as in Gaspar & Rieke 2020). We simulated noise-injected companion PSFs to estimate the random uncertainties of the companion centroids in the CHARIS images, providing an empirically-motivated estimate of our centroid uncertainties. The simulated PSFs were made by adding noise floors to the forward-modeled PSFs (see Section 2.1 and Currie et al. 2018). The noise floors were taken from the areas in the final images created by combining the PSF-subtracted cube frames. Then, we used the areas at the same separations as the companion but the different 12 PAs, which starts from the companion's PA + 45 • and ends at the companion's PA + 320 • with intervals of 25 • . The standard deviations of the centroids calculated from the simulated PSFs were taken to be the random uncertainties of the companion centroid measurements. ) for the 2020-October, 2020-December, and 2021-October CHARIS images. We do not perform the same analysis for the NIRC2 L data since the SNR is lower and intrinsic PSF is roughly twice as large as CHARIS's (θ ≈ 0. 08 vs. 0. 043). In this case, the centroid uncertainty estimated from the intrinsic SNR is ∼ 0.35 pixels: significantly larger than for any CHARIS measurement.
Another source of the random errors is attributed to the alignment of the individual images. We evaluated this error source using the SCExAO/CHARIS data sets obtained in October and December 2020 for HIP 21152. The residuals between the central star's positions calculated at each wavelength channel of a cube and the polynomial function fit to the positions can correspond to the image alignment errors. The residual scatter of an image alignment is much smaller than the other insignificant error sources (see above), and further decreases when integrating all frames and all channels; we thus neglect the image alignment errors in SCExAO/CHARIS. Table 4 summarizes the error evaluations described above for the CHARIS measurements. When the astrometric calibrations for SCExAO/CHARIS will be updated in future (e.g., for distortion calibration), we recommend the reader to refer to Table 4 for recalculating the astrometric measurements with the updated calibrations.

D.2. Common Proper Motion and Orbit Analysis
In Figure 7, we compare the measured positions of HIP 21152 B relative to HIP 21152A with the positions expected if HIP 21152 B is an unbound background object (i.e., common proper motion analysis). The expectation was made with HIP 21152A's right ascension (RA), declination (DEC), RA and DEC proper motions, and parallax from Gaia EDR3 (Gaia Collaboration et al. 2021). The results from our orvara orbit modeling for the HIP 21152 system (Section 4.2) are summarized in Table 5. In addition to Figure 4, Figure 8 shows the fitted orbits to HIP 21152A's proper motion variations from HGCA (Brandt 2021) along RA and DEC and HIP 21152 B's projected separations and position angles.

E. SUBSTELLAR COMPANION CANDIDATES AROUND MAIN-SEQUENCE STARS IN THE HYADES
There have been no previous reports of confirmed substellar companions to main-sequence stars in the Hyades open cluster (OC), despite its proximity to the solar system. We note that Morzinski et al. (2012) reported candidates of Hyades substellar companions detected with shallow adaptive optics imaging. In addition, an L1 ± 1 type companion around the M4-type star 1RXS J034231.8+121622 was reported by Bowler et al. (2015). A similarity of the space velocities and the sky position of this system with those of the Hyades OC was already discussed in Bowler et al. (2015); nevertheless, they concluded that this system's membership to the Hyades OC is unclear. For 1RXS J034231.8+121622, we run the Banyan-Σ algorithm (Gagné et al. 2018) with the star's distance adopted in Bowler et al. (2015), RA, DEC, and proper-motion measurements from Gaia Collaboration et al. (2021), and absolute RV from Shkolnik et al. (2012), providing a zero probability as a Hyades member. In contrast, we compute a membership probability for the same star of 99.9% using the same algorithm after updating the distance to a Gaia-based measurement from Bailer-Jones et al.
; the different membership probability is therefore caused by Gaia data becoming available. Although 1RXS J034231.8+121622 B can be identified as a high-probability member of the Hyades OC, it is still unclear whether the companion has a substellar mass. We calculate a mass of 1RXS J034231.8+121622 B with the new membership and the Gaia-based distance measurement. For this companion, we adopt a distance of 32.96 +0.022 −0.024 pc from Bailer-Jones et al. (2021) and calculate an H-band bolometric correction of 2.70 ± 0.08 mag following Liu et al. (2010) to convert an H-band apparent magnitude of 13.51 ± 0.05 (Bowler et al. 2015) to the bolometric luminosity of log (L/L ) = −3.552 ± 0.038. With the age of Hyades (750 ± 100 Myr) and the evolutionary models of Baraffe et al. (2003), the bolometric luminosity is converted to a mass of 76-83 M Jup , which is near or slightly above the hydrogen burning limit (≈70-80 M Jup in general; e.g., Fernandes et al. 2019). We thus find that 1RXS J034231.8+121622 B is a candidate substellar companion in the Hyades OC.