First Observations of the Brown Dwarf HD 19467 B with JWST

NIRCam observed HD 19467 on 2022 August 12 with the long-wavelength bar (LWB) coronagraphic mask in subarray mode with six filters: F250M, F300M, F360M, F410M, F430M, and F460M. The target star was observed at two telescope roll angles separated by 772. Table 1 shows a summary of the observations and settings per filter. Observations of HD 19467 were taken with the long-wavelength bar coronagraph mask (MASKLWB), providing a test of NIRCam's capabilities at smaller inner working angles than are possible with the round masks (4λ/D for MASKLWB versus 6λ/D for MASK210R, MASK335R, and MASK430R; Krist et al. 2007). At the time of these observations, the MASKLWB positions were not well-defined, with a y-offset ∼70 mas. Future use of this mode with an updated position definition will improve the ability to center the star on the mask and therefore contrast performance, especially close in to the mask. These observations represent one of the first post-commissioning uses of the bar coronagraph.

The observation plan was initially scheduled to include sequential observations of the reference star HD 19096 in order to perform point-spread function (PSF) subtraction using reference differential imaging (RDI). However, the reference observations were unsuccessful because the telescope failed to acquire a guide star. Instead, we performed post-processing using only angular diversity along with models of the telescope's and instrument's optical performance enabled by regular measurements of the telescope wave front error, simulating reference star differential imaging but without the actual observation of a reference star. A very similar approach has been applied to enable high-contrast imaging with the Hubble Space Telescope by modeling the instrumental PSF (e.g., Krist et al. 1997); JWST's stability and regular measurements of the wave front further enable this technique. High-contrast observations with only angular diversity can significantly reduce the observation time and overhead. We demonstrate that this can be an appropriate strategy for bright and widely separated companions.

New RV measurements of HD 19467 were obtained in 2022 July through 2022 August using the High Resolution Echelle Spectrometer (HIRES) on the Keck I Telescope. The new RV measurements are processed using standard data reduction techniques described in Butler et al. (1996) and Butler et al. (2017). The majority of the RVs come from Rosenthal et al. (2021), where the reduction techniques are described in more detail. In brief, the HIRES RV values are measured using an iodine cell-based design in order to calibrate the wavelengths in the stellar spectrum. The spectral region from 5000 to 6200 Å is used for measuring the RVs. We combine the new observations with previous measurements for a total of 53 RV measurements spanning 25 yr for the data analysis. The data including the new measurements are listed in Table 8 in Appendix B.

HD 19467 was observed by the TESS spacecraft (Ricker et al. 2015) in Sectors 4 and 31, resulting in ≈60 days of high-precision optical photometry. Sector 31 includes data obtained with a 20 s cadence, a new observing mode introduced in the TESS extended mission. The TESS 20 s data show improved photometric precision for bright stars such as HD 19467 (Huber et al. 2022), and we therefore focus on the 20 s data here. We used the PDC-MAP light curves provided by the Science Processing Operations Center (SPOC; Jenkins et al. 2016), which have been optimized to remove instrumental variability (Smith et al. 2012; Stumpe et al. 2012), and remove all data with quality flags not equal to zero, which yields the best precision of the 20 s data (Huber et al. 2022).

The top panel of Figure 1 shows the TESS 20 s cadence light curve for HD 19467. We observe no significant long-term variability, with an rms of 25.5 ppm over 6 hr timescales. To search for high-frequency variability, we used the established asteroseismic tools pySYD (Huber et al. 2009; Chontos et al. 2021) and FAMED (Corsaro et al. 2020), which analyze the data in the frequency domain. Both methods detected a significant power excess near ≈2200 μHz, consistent with the expected ≈7 minute timescale of solar-like oscillations (Bedding 2014; García & Ballot 2019) based on the spectroscopic temperature and surface gravity (Table 2). We also analyzed the data in the time-domain using a Gaussian process (GP) model with a stochastically driven damped harmonic oscillator (Foreman-Mackey et al. 2017), which has been demonstrated to outperform traditional frequency analysis tools in recovering low signal-to-noise (S/N) oscillations (D. R. Hey et al. 2023, in preparation). The GP analysis strongly favored a model with an oscillating component with a Bayesian information criterion (BIC) of ΔBIC = 7.1.

Zoom In Zoom Out Reset image size

The bottom panel of Figure 1 shows a power spectrum of the 20 s light curve centered on the power excess. Solar-like oscillations are described by a frequency of maximum power (${\nu }_{{\rm{\max }}}$) and a large frequency separation (Δν), which approximately scale with $\mathrm{log}g$ and the mean stellar density, respectively (Ulrich 1986; Brown et al. 1991). We derive ${\nu }_{{\rm{\max }}}=2180\pm 100$ μHz, with the central value taken from the median of three solutions (pySYD, FAMED, and GP), and uncertainties calculated from the scatter over individual methods (e.g., Huber et al. 2013). The low S/N of the detection precludes an unambiguous detection of Δν. Visual inspection of the echelle diagram indicates Δν ≈ 101 μHz, consistent with the derived ${\nu }_{{\rm{\max }}}$ value.

We adopted the effective temperature (T_eff) and metallicity ([M/H]) from Brewer et al. (2016), derived from a line-by-line analysis of a Keck/HIRES spectrum. Literature values from spectroscopy and Gaia color–temperature relations (Casagrande et al. 2021) are highly consistent, with a range of 40 K in T_eff and 0.04 dex in iron abundance (Table 2). We used these ranges as estimates for the uncertainties, resulting in T_eff = 5747 ± 40 K and [M/H] = −0.09 ± 0.04 dex. These uncertainties are smaller than those recommended by Tayar et al. (2022), which are justified by the fact that the star has properties similar to the Sun and thus suffers from smaller systematic errors.

We then combined the asteroseismic ${\nu }_{{\rm{\max }}}$ measurement, Gaia DR3 parallax, 2MASS K-band magnitude, T_eff, and [M/H] with isoclassify (Huber et al. 2017) and BASTA (Aguirre Børsen-Koch et al. 2022), which perform Bayesian inference of stellar parameters given input observables using the stellar evolution models MIST (Choi et al. 2016) and BASTI (Pietrinferni et al. 2004), respectively. Importantly, ${\nu }_{{\rm{\max }}}$ tightly constrains the surface gravity to $\mathrm{log}g=4.28$, which combined with the radius constraint from the Gaia parallax provides a tight constraint on its stellar mass, which in turn constrains its stellar age. Both tools consistently imply a mass of ≈0.95 M_☉, which, given that the star has slightly evolved off the main sequence (1.2 R_☉), implies an old age. Figure 2 shows the age posteriors from both evolutionary models and methods.

Zoom In Zoom Out Reset image size

The final stellar parameters adopted in our study are listed in Table 3. We adopt the self-consistent solution derived from isoclassify, but add in quadrature the difference to the BASTA results to account for systematic errors due to different model grids (Tayar et al. 2022).

Table 3. Adopted Stellar Parameters for HD 19467A

Effective temperature, Teff (K)	5747 ± 40
Metallicity, [M/H] (dex)	−0.09 ± 0.04
Luminosity, L (L☉)	1.42 ± 0.06
Stellar radius, R⋆ (R☉)	1.20 ± 0.03
Stellar mass, M⋆ (M☉)	0.96 ± 0.02
Stellar density, ρ⋆ (cgs)	0.56 ± 0.03
Surface gravity, $\mathrm{log}g$ (cgs)	4.28 ± 0.04
Age, t (Gyr)	9.4 ± 1.0

Note. T_eff and [M/H] are adopted from Brewer et al. (2016), with uncertainties accounting for the spread in literature results. All other properties are derived from the combination of constraints from asteroseismology, spectroscopy, and Gaia data (see Section 3).

Download table as:  ASCIITypeset image

The age of HD 19467 is important for interpreting the mass and atmospheric composition of the brown dwarf companion. As already mentioned, literature estimates have a significant spread, ranging from 5 to 12 Gyr (Table 3). Younger, Sun-like ages come from stellar rotation (Maire et al. 2020) and activity (Brandt et al. 2021a), while older ages are preferred by isochrone fitting (Wood et al. 2019; Maire et al. 2020). The rotation-age is based on the detection of a photometric period of ≈29 days with an amplitude of ≈0.5% from ground-based ASAS data (Maire et al. 2020). The high-precision of the TESS light curve in Figure 1 rules out rotational modulation at the level of 0.5% over the 29 day timescale suggested by the ASAS data, which implies that the rotation period for HD 19467 is undetermined. This is consistent with results from the Kepler Mission, which demonstrated that typical rotational amplitudes in mature Sun-like stars are on the order of a few hundred ppm (McQuillan et al. 2014; Santos et al. 2021) and thus are generally not detectable using ground-based photometry. Chromospheric activity-based ages also become more challenging for stars with Sun-like and older ages due flattening of the age–activity relation, making age constraints sensitive to small changes in $R{{\prime} }_{\mathrm{HK}}$ measurements. While some literature values for $R{{\prime} }_{\mathrm{HK}}$ favor near-solar values (and thus ages) for HD 19467, others are consistent with older, isochrone-based ages (e.g., $R{{\prime} }_{\mathrm{HK}}=5.1$ and 8.8 ± 0.3 Gyr; Lorenzo-Oliveira et al. 2018).

The asteroseismic detection from TESS supports an older age for HD 19467. While the low S/N precludes a direct age from a measurement of individual oscillation frequencies (e.g., Mathur et al. 2012; Metcalfe et al. 2014; Silva Aguirre et al. 2017), the ${\nu }_{{\rm{\max }}}$ measurement precisely constrains $\mathrm{log}g$ and thus stellar mass independent of stellar evolutionary models. With a mass similar to solar (0.96 ± 0.02 M_☉), HD 19467 must have an age significantly older than the Sun to reach a radius of 1.2 R_☉. ¹⁶ As discussed by Maire et al. (2020), an age of 9.4 ± 1.0 Gyr is compatible with the slight enhancements in alpha elements and subsolar metallicity, placing the star in the transition region between chemical "thin-disk" and "thick-disk" stars. Overall, we conclude that HD 19467 is an ≈4–5 Gyr older analog to our Sun and adopt an age of 9.4 ± 1.0 Gyr (Table 3).

We apply principal component analysis (PCA; e.g., Lafreniere et al. 2007; Amara & Quanz 2012) via Karhunen Loéve Image Projection (KLIP; Soummer et al. 2012), to subtract the residual stellar intensity from the science frames using the images taken at two roll angles for angular diversity. We perform PSF subtraction using the open source Python package pyKLIP (Wang et al. 2015) using angular differential imaging (ADI) and RDI using a synthetic reference PSF, as described below. The results of the PCA reduction for all filters are displayed in Figure 3.

Zoom In Zoom Out Reset image size

For all filters except F250M, the data contained in the two rolls suffice to obtain an unambiguous detection of the companion. For the F250M case, although the companion's signal is visible using only ADI, we resorted to using RDI with a set of synthetic PSFs in order to confirm that the signal is indeed from the companion and not due to residual speckles. In Figure 4 we show a comparison, for the F250M filter, between only using the roll frames for the PCA reduction (ADI), and assisting the PCA reduction with a set of synthetic PSFs (ADI+SynRDI).

Zoom In Zoom Out Reset image size

A grid of synthetic stellar PSFs is generated using WebbPSF (Perrin et al. 2014) and tools from webbpsf_ext ¹⁹ at offset locations with respect to the coronagraph focal plane mask. We generate simulated PSFs in different sets of 9-point grid patterns at even spacings. We simulate spacings of 2.5, 7, 15, 25, and 40 mas, in addition to a set of rotations of the coronagraphic-PSF with respect to the detector of 01, 03, and 05. This aims to emulate the speckles present in the data frames, and assists the PCA reduction with the diversity in speckle structure needed to perform a more optimal reference subtraction.

As mentioned above, for filters F300M, F360M, F410M, F430M, and F460M, ADI suffices for a clear detection. As a second step we use RDI with the synthetic PSFs to subtract the unwanted starlight further. This is motivated by the fact that the companion PSF's northern lobe falls near the diffraction speckles caused by the bar coronagraph. The number of KL modes determines how many of the synthetic PSFs are used for the subtraction. Since these have been generated with arbitrary offsets, there is a risk of subtracting light from the secondary. We use 15 KL modes, which minimizes over-subtraction and clears out slightly more of the residual starlight around the northern lobe. This was done by visual inspection; a more in-depth analysis on how to use synthetic PSFs optimally will be explored in the future.

To extract the flux and position of HD 19467 B accurately, we account for over-subtraction effects on the PSF that arise during the reduction process (described in Section 4.1) with a forward model based on the method described by Pueyo (2016). We make use of its implementation on pyKLIP (Wang et al. 2015). The companion PSF is modeled using WebbPSF (Oschmann et al. 2014) for each filter and accounting for its position with respect to the bar focal plane mask. An accurate position of the simulated PSF is particularly important in the case of the shorter wavelength filters: poorer spatial sampling of the pixels compared to the diffraction limit at smaller wavelengths (2.5 μm is sub-Nyquist) results in an acute sensitivity of the PSF structure as seen in the detector. Accurate positioning for the case of F250M was done by trial and error simulating a grid of PSF offsets and selecting the best fit by the least-squares difference between the simulated and the coadded science frames. The model PSFs are simulated using the OPD map closest in time and prior to the observations.

The flux and position of the companion are extracted with pyKLIP. We fit a model of its photometry and astrometry to the reduced data using an Markov Chain Monte Carlo (MCMC) approach (emcee; Foreman-Mackey et al. 2013). Figure 5 shows the PSF model fit to the reduced data for F360M, the filter in which we obtain the highest S/N. Appendix A contains the full gallery of forward-model comparisons with the PSF-subtracted images in each band.

Zoom In Zoom Out Reset image size

The flux calibration of the signal is determined based on a jwst stage 2 pipeline photometric calibration. We apply a flux correction to the photometry based on a measured attenuation factor of 0.92 of the bar mask Lyot stop at ∼16. We fit a G3V stellar photosphere model to 1–5 μm photometry from 2MASS (Cutri et al. 2003) and WISE (Cutri et al. 2012). We also find a ∼2% error in fitting the stellar model to the IR measurements, and apply this error to contrast the reported values. Table 4 shows the estimated stellar flux. Comparison of the calibrated flux measured from the acquisition and astrometric confirmation images, both taken through the neutral density square, produced from the stage 2 pipeline is consistent with the estimated stellar spectrum within ∼10%. We therefore apply a 10% uncertainty to the reported absolute photometry of HD 19467 B in this section.

Figure 6 shows the measured photometry in each NIRCam band alongside recent measurements and limits from the ground (Maire et al. 2020; Mesa et al. 2020). The F460M flux is consistent with the M-band upper limit obtained with the Very Large Telescope (VLT) Nasmyth Adaptive Optics System Near-Infrared Imager and Spectrograph (NACO; Maire et al. 2020), however there appears to be some tension with the NACO L' flux compared with the F360M and F410M photometry measurements. The difference in passbands on a steeply rising part of the spectrum, possible water vapor effects, as well as calibration uncertainties may account for this discrepancy. Continued refinement of JWST's photometric calibrations will help identify any biases in the photometry. For this study, we do not incorporate NACO photometry into the analysis, but will rather present a measurement comparison in a future investigation.

Zoom In Zoom Out Reset image size

Table 5 shows our measured photometry and relative astrometry for HD 19467 B (see the next section).

A major challenge in obtaining relative astrometry of a companion in coronagraphic imaging is that the primary star is occulted by the focal plane mask. Knowledge of the wave front from published OPD maps and a highly structured PSF enable a forward-model-based cross-correlation with the data to fit for the centroid of the star behind the mask. We perform a cross-correlation of the model PSFs with the data using chi2_shift in the image-registration Python package ²⁰ to measure the best-fit position of the star behind the mask (Figure 7). We obtain a centroiding error of ∼7 mas, consistent with the measured sensitivity in Carter et al. (2022).

Zoom In Zoom Out Reset image size

The companion position is recovered with the joint astrometry and photometry model fit to the reduced data as described in Section 4.2. The model fit errors provide the uncertainty in the relative position to the measured star position on the detector. We add the star position uncertainties to the reported errors (Table 5). Figure 8 shows the new astrometric measurement compared to previous relative astrometry measurements of HD 19467 B (Crepp et al. 2014, 2015; Bowler et al. 2020; Maire et al. 2020).

Zoom In Zoom Out Reset image size

In Figure 9 we show contrast curves for the reduced images after PSF subraction. The contrast is measured with PyKLIP by computing the noise in an azimuthal annulus at each separation, using a Gaussian cross-correlation to remove high-frequency noise. The flux normalization to obtain these contrast numbers was computed as explained in Section 4.2, by using a best-fit model of the stellar spectrum to calibrate the contrast. The contrast curves are corrected for algorithmic throughput, i.e., throughput loss due to the PSF subtraction, and for small sample statistics (Mawet et al. 2014).

Zoom In Zoom Out Reset image size

Figure 10 translates our detection limits from flux/contrast sensitivities to limits on companion mass. Three different brown dwarf evolution models are considered: Ames-COND (Baraffe et al. 2003), BEX-HELIOS (Linder et al. 2019), and Sonora-Bobcat (Marley et al. 2021). In each case, we assume solar metallicity.

Zoom In Zoom Out Reset image size

While the shortest wavelength observations achieve the best contrast (in particular F300M), the longer wavelengths are better at detecting lower-mass companions. We find an overall detection limit of ∼10 M_J. This limit is much higher than the sub-Jupiter levels that JWST/NIRCam can obtain for young systems (e.g., Carter et al. 2022), but even for this very old system brown dwarfs are easily detectable outside of ∼04 ≃ 10 au. While the detection limit is relatively independent of the model, it does depend significantly on the age of the brown dwarf (the system age is discussed in Section 3.3).

Despite a lack of reference star observations, we are able to recover the signal of HD 19467 B with two roll angles and achieve contrasts of ∼10⁻⁵ at 1''–2''. Regular OPD measurements enable the use of synthetic PSFs that can aid PSF subtraction by generating a set of reference PSFs to capture speckle structure. This suggests that bright companions could be observed without reference stars, significantly reducing the time spent on the observation. Future work will investigate the difference between reducing data with and without reference star observations. Future observations with a better defined position for the LWB coronagraph should also provide better contrast close in.

We compare the NIRCam photometry with both the Sonora-Bobcat and Sonora-Cholla model grids to highlight the broad features of the 2–5 μm flux. We allow the radius scaling to vary when fitting the models to our NIRCam photometry, which is generally consistent with radii much smaller than, e.g., those predicted by the Sonora-Bobcat grid evolutionary tables.

We find that the Bobcat models do not simultaneously capture both the local flux peak at 3 μm and the steep drop in flux from 4 to 5 μm. In general the Bobcat grid favors higher gravity and lower or zero metallicity. However, none of the fits capture all of the photometry perfectly. A model spectrum at an effective temperature of T_eff = 1400 K best matches the data, but does not fully capture the peak flux at ∼4 μm with the drop off at longer wavelengths. The gravity and metallicity do not have a strong effect on the fit.

The Sonora-Cholla models, which include effects of disequilibrium chemistry, provide better fits to the photometry, but suggest a low value for the gravity as well as a small radius to scale the flux. The best-fit model has T_eff = 1200 K, g = 31, and ${log}({K}_{{zz}})=2$. Figure 15 (right) shows the best-fitting Cholla model, with a few model grid points that vary in temperature, eddy diffusion parameter, and gravity. We note that while lower gravity is favored, the gravity does not strongly affect the shape of the photometry-only profile, as can be seen in the third panel. Near-IR spectroscopy with JWST will be able to distinguish the features of gravity further.

Zoom In Zoom Out Reset image size

Overall, the best-fit Sonora-Cholla model provides a better fit than the best-fit Sonora-Bobcat model. As can be seen in Figure 15, the Cholla models are better able to capture the lower flux at shorter wavelengths simultaneously, the peak at ∼4 μm, and the subsequent drop in flux, favoring a lower temperature that is more consistent with previous estimates. This suggests that modeling disequilibrium chemistry may be necessary for understanding the atmosphere and evolution of HD 19467 B.

We fit the Sonora-Cholla cloudless models to a composite data set that consists of JWST photometry, IRDIS spectra (Mesa et al. 2020), and SPHERE K12 band photometry (Maire et al. 2020) to constrain the temperature and gravity of HD 19487 B. For the IRDIS spectra, we exclude spectral points with negative flux values. We include an additional uncertainty of 7%, similar to the J- and H-band absolute flux uncertainties of HD 19467 (Table 3 of Mesa et al. 2020), to account for the possible offset in the absolute flux levels.

We linearly interpolate the Cholla model spectral grid to a finer grid with smaller step sizes in the temperature, gravity, and eddy diffusion parameter. In addition to these three parameters, the other free parameter in the spectral fitting is the scaling factor, which is the square of the ratio of brown dwarf radius to the distance (32.03 pc). We use pyphot ²² to calculate the broadband photometry of the model spectra.

Our model fitting algorithm minimizes a cost function that sums the squared difference between the model spectra and data, weighted by the observational uncertainties, σ, and the intensities, w, as shown in the following:

$\begin{eqnarray}\begin{array}{rcl}{\rm{cost}}\,{\rm{function}} & = & \displaystyle \sum _{i\,=\,1}^{N}{\left\{{w}_{i}\displaystyle \frac{{F}_{\lambda ,i}({\rm{model}})-{F}_{\lambda ,i}({\rm{data}})}{{\sigma }_{i}}\right\}}^{2},\text{where},\\ & & {w}_{i}=\displaystyle \frac{{F}_{\lambda ,i}{\rm{\Delta }}{\lambda }_{i}}{max({F}_{\lambda ,i}{\rm{\Delta }}{\lambda }_{i})},\end{array}\end{eqnarray} \tag{ 1 }$

where Δλ_i is the wavelength coverage of a data point and the intensity weighting is normalized by the highest intensity among the data points. We include the intensity weighting in the cost function so that a photometric point with high intensity carries more weight in the fitting process than a spectral point with low intensity even though both could have a similar S/N, since the photometric points contain a larger fraction of the total flux measured. We then use emcee (Foreman-Mackey et al. 2017) to sample the posterior distribution of the fitted parameters with the MCMC method. We adopt uniform priors of the temperature, logarithmic gravity, and eddy diffusion parameter. We run the MCMC chain with 200 walkers for 20,000 steps. Based on the posterior distribution of the MCMC chains, we derive a best-fit temperature of 1080 ± 22 K, gravity of $\mathrm{log}(g)={4.60}_{-0.1}^{+0.2},$ eddy diffusion parameter of $\mathrm{log}{K}_{\mathrm{zz}}={3.1}_{-0.4}^{+0.6}$, and radius R of 0.62 ± 0.03 R_J. The value for the radius is lower than the expected radius at 9 Gyr, ∼0.8 R_Jup, according to the evolutionary model grid at similar parameters. Prior modeling work has noted a common discrepancy between the expected radius from evolutionary models and the radius required to match the flux of a given temperature that best fits the atmospheric model (e.g., Barman et al. 2011; Marley et al. 2012; Lavie et al. 2017). Maire et al. (2020) explored radius agreement with evolutionary tracks with different model atmospheres that included various levels of clouds in the atmospheres. Future observations, especially spectroscopy will further help constrain atmospheric models.

We draw 100 parameter sets from the posterior distribution and plot the corresponding model spectra in Figure 16. Figure 16 suggests that the model spectra provide a qualitatively good match to the data but there are some significant residuals, especially in the K-band photometry. We remain cautious about the inferred parameters and the seemingly small uncertainties given the imperfect fit between the data and model spectra. However, while the absolute flux calibrations may account for some of the discrepancy between the data sources, the discrepancy alone is not enough to account for inconsistencies between the best-fit atmospheric model and evolutionary grid predictions. Clouds, which likely drive the rotational modulation of many T dwarfs (e.g Manjavacas et al. 2019) and are not included in the Cholla models, could play a key role in shaping the HD 19467 B emission spectra. Future simultaneous observation in the near-IR and mid-IR regions will be useful for testing the role of clouds in the atmosphere of HD 19467 B with well-constrained age and host-star metallicity.

Zoom In Zoom Out Reset image size

The 2–5 μm JWST NIRCam broadband photometry, in combination with the ground-based near-IR spectra and photometry, are crucial for pinning down the bolometric luminosity of a ∼1000 K object. We integrate the flux density over the observed data including the previously published IRDIS-LSS spectra (0.97–1.335 μm and 1.50–1.80 μm) , ground-based K-band photometry (2.059–2.161 μm and 2.1965–2.3055 μm), and the six JWST NIRCam broadband photometry. The flux integral in the observed wavelength regions is (3.28 ± 0.2) × 10⁻⁶ L_⊙.

We utilize the fitted model spectra in Section 6 to extrapolate flux density beyond the observed wavelength region and estimate the bolometric luminosity. We find that the observational data account for around 72% of the bolometric luminosity. The estimated total bolometric luminosity is (4.75 ± 0.2) × 10⁻⁶ L_⊙, or $\mathrm{log}(L/{L}_{\odot })=-5.32\pm 0.02$. Based on the combination of JWST NIRCam high-precision photometry and ground-based data, our results suggest that we can accurately derive the bolometric luminosity at a 4% precision level.

Based on the independently estimated age and bolometric luminosity, we then use the Bobcat evolution model to estimate the mass of HD 19467 B. After linearly interpolating the mass as a function of age and bolometric luminosity, we derive that the mass is 62 ± 1 M_J . Figure 17 shows the Sonora-Bobcat evolution model and where our bolometric luminosity estimate lies. By comparing the mass derived from the age and L_bol to the dynamical mass, (${81}_{-12}^{+14}{M}_{J}$), we conclude that the two masses are consistent with each other within about 2σ.

Zoom In Zoom Out Reset image size

The NIRSpec IFU observations planned for later this year (PID #1414) will provide a much more complete characterization of the atmosphere of HD 19467 B, helping to pin down parameters such as metallicity and T_eff.