JWST Observations of the Enigmatic Y Dwarf WISE 1828+2650: I. Limits to a Binary Companion

TL;DR

This study uses JWST and Keck observations to investigate whether the cold Y dwarf WISE 1828+2650 has a binary companion, finding no evidence for multiplicity beyond 0.5 AU, which impacts models of its physical characteristics.

Contribution

The paper provides new high-resolution imaging data that constrains the presence of a binary companion to WISE 1828+2650, refining understanding of its nature and aiding future spectral modeling.

Findings

No evidence for a binary companion beyond 0.5 AU from JWST and Keck data.

New photometry from JWST in 6 filters for improved model fitting.

Constraints on the multiplicity of the Y dwarf WISE 1828+2650.

Abstract

The Y-dwarf WISE 1828+2650 is one of the coldest known Brown Dwarfs with an effective temperature of $\sim$300 K. Located at a distance of just 10 pc, previous model-based estimates suggest WISE1828+2650 has a mass of $\sim$5-10 Mj, making it a valuable laboratory for understanding the formation, evolution and physical characteristics of gas giant planets. However, previous photometry and spectroscopy have presented a puzzle with the near-impossibility of simultaneously fitting both the short (0.9-2.0 microns) and long wavelength (3-5 microns) data. A potential solution to this problem has been the suggestion that WISE 1828+2650 is a binary system whose composite spectrum might provide a better match to the data. Alternatively, new models being developed to fit JWST/NIRSpec and MIRI spectroscopy might provide new insights. This article describes JWST/NIRCam observations of WISE 1828+2650 in 6 filters to address the binarity question and to provide new photometry to be used in model fitting. We also report Adaptive Optics imaging with the Keck 10 m telescope. We find no evidence for multiplicity for a companion beyond 0.5 AU with either JWST or Keck. Companion articles will present low and high resolution spectra of WISE 1828+2650 obtained with both NIRSpec and MIRI.

Figures

Tables

Table 1. Table 1: JWST NIRCam Observing Parameters (PID:#1189)

Instrument	Filter Pair	Readout	Groups/Int	Ints/Exp	Dithers	Total Time (sec)
NIRCam	F115W & F335M (CH4)	MEDIUM	4	10	4	1632
NIRCam	F140M (H2O, CH4) & F360M	MEDIUM	4	10	4	1632
NIRCam	F162M (ref for F140M) & F470N (H2)	MEDIUM	4	10	4	1632

Table 2. Table 2: NIRCam Photometry

Filter	F${}_{\nu }{}^{}(\mu$Jy)1	Magnitude
F090W	0.078$\pm$0.005(0.015)	26.16$\pm$0.20
F115W	0.311$\pm$0.005(0.03)	24.39$\pm$0.10
F162M	1.29$\pm$0.009(0.13)	22.25$\pm$0.10
F335M	2.41$\pm$0.013(0.24)	20.23$\pm$0.10
F360M	34.2$\pm$0.05(3.4)	17.19$\pm$0.10
F470M	340.4$\pm$0.62(34)	14.16$\pm$0.10
F480M2	326$\pm$0.14(32)	14.21$\pm$0.10
J (1.25 $\mu$m)	0.65$\pm$0.23	23.48$\pm$0.23
H (1.65 $\mu$m)	0.83$\pm$0.21	22.73$\pm$0.13
K (2.16 $\mu$m)	0.27$\pm$0.08	23.48$\pm$0.36
Spitzer/IRAC (3.55 $\mu$m)	48.1$\pm$1.0	16.92$\pm$0.02
Spitzer/IRAC (4.55 $\mu$m)	335.5$\pm$6.71	14.32$\pm$0.02
WISE W2 (4.62 $\mu$m)	310$\pm$14	14.35$\pm$0.05

Table 3. Table 3: NIRCam Astrometry of WISE 1828

Observatory	Epoch	RA	Dec	Parallax	${\mu }_{RA}$	${\mu }_{DEC}$
	(MJD)	RA (Equinox=J2000)	(Equinox=J2000)	(mas)	(mas/yr)	(mas/yr)
Spitzer1	57094.09	277.131096 ($\pm$2 mas)	26.844069($\pm$2 mas)	100.3$\pm$2	1016.5$\pm$0.8	169.3$\pm$0.8
		18h28m31.463s	+26o50′38.65″
JWST (predicted)	59788.5143	277.1334156 ($\pm$6 mas)	26.8444348 ($\pm$6 mas)
		18h28m32.020s	+26o50′39.965″
JWST (F360M)	59788.5143	277.1334164 ($\pm$10 mas)	26.8444334 ($\pm$10 mas)
		18h28m32.020s	+26o50′39.960″
$\Delta$Spitzer-JWST(pred)		$-$7.45″	$-$1.32″
$\Delta$JWST (Pred-Obs)		2.5$\pm$12 mas	-5.1$\pm$8.2 mas
Gaia-JWST differences3 (mas)		10.4$\pm$10.2 (1.8)	6.8$\pm$5.1 (0.9)

Table 4. Table 4: Spectral Model Fits to Photometry

		Primary			Secondary
	Teff	log g	Mass	Radius	Teff	log g	Mass	Radius	Metallicity	Age	Reduced	Degs. of	$\log$ ($L/L_{\odot }$)
	(K)	(cgs)	(MJup)	(RJup )	(K)	(cgs)	(MJup)	(RJup)	[M/H]	($\times {10}^9$yr)	${\chi }^2$	Freedom
Sonora Cloudless models
best-fit single	325	3.6	2	1.83					-0.5∗	0.3	180	9	-6.49$\pm$ 0.03
99.7$\%$ CI	309–405	3.0–5.5	0.6–13	1.1–2.0					-0.5	0.01–7
best-fit binary	337	4.0	4.4	1.38	330	4.0	4.1	0.90	-0.5∗	0.7	231	7	-6.49$\pm$0.04
99.7$\%$ CI	316–413	3.0–4.8	0.6–20	0.4–2	268–395	3.5–4.8	0.6–20	0–2	-0.5	0.03–10
ATMO Chemical Equilibrium models
best-fit single	300	4.0	3.7	2					0 (fixed)	1.0	256	10	-6.45$\pm$ 0.06
99.7$\%$ CI	287–359	3.0–5.1	0.5–13	2–2					-	0.02–8.5
best-fit binary	300	3.6	1.5	1.9	300	3.6	1.7	1.4	0 (fixed)	0.3	287	8	-6.36$\pm$0.05
99.7$\%$ CI	279–400	3.0–4.8	0.5–20	0.4–2	279–400	3.0–4.8	0.5–20	0–2	-	0.02–10
ATMO Chemical Non-equilibrium models with strong mixing
best-fit single	294	4.3	5.5	2.0					0 (fixed)	2.4	303	10	-6.46$\pm$ 0.06
99.7$\%$ CI	259–359	3.0–5.0	0.5–11	2.0–2.0					-	0.02–9.9
best-fit binary	270	4.0	4.2	2	270	4.0	4.2	2	0 (fixed)	2.1	320	8	-6.38$\pm$0.09
99.7$\%$ CI	253–340	3.0–4.6	0.5–13	2–2	253–340	3.0–4.6	0.5–13	2–2	-	0.03–9.9

Table 5. Table 5: Highly Red Sources in WISE 1828+2650 Field

Label	RA20001	DEC20001	F090W2	F115W2	F162M2	F335M2	F360M2	F470N2	FWHM3
1	277.138408(60.)	26.8286435(55.)	1.06$\pm$0.00769	4.08$\pm$0.00955	0.911$\pm$0.00908	6.23$\pm$0.0223	5.66$\pm$0.0218	4.24$\pm$0.0982	0.15″
	18h28m33.22s	26d49m43.12s	23.33$\pm$0.0078	21.59$\pm$0.0025	22.62$\pm$0.011	19.2$\pm$0.0039	19.15$\pm$0.0042	18.92$\pm$0.025
2	277.160802(20.)	26.8420123(16.)	0.172$\pm$0.00651	0.545$\pm$0.00633	0.975$\pm$0.00975	2.$\pm$0.0113	2.27$\pm$0.0122	2.5$\pm$0.0747	0.23″
	18h28m38.59s	26d50m31.24s	25.3$\pm$0.04	23.78$\pm$0.013	22.55$\pm$0.011	20.43$\pm$0.0061	20.14$\pm$0.0058	19.5$\pm$0.032
34	277.133408(26.)	26.8444326(4.3)	0.078$\pm$0.005	0.311$\pm$0.00446	1.29$\pm$0.00857	2.41$\pm$0.0132	34.2$\pm$0.0511	340.$\pm$0.619	0.15″
	18h28m32.02s	26d50m39.96s	26.16$\pm$0.20	24.39$\pm$0.015	22.25$\pm$0.0072	20.23$\pm$0.0059	17.19$\pm$0.0016	14.16$\pm$0.002
4	277.16172(24.)	26.8470549(26.)	0.0796$\pm$0.00708	0.323$\pm$0.00663	0.562$\pm$0.00985	1.93$\pm$0.0112	2.14$\pm$0.012	2.55$\pm$0.0752	0.44″
	18h28m38.81s	26d50m49.4s	26.14$\pm$0.093	24.34$\pm$0.022	23.15$\pm$0.019	20.47$\pm$0.0063	20.2$\pm$0.0061	19.48$\pm$0.032
5	277.138946(24.)	26.8479253(5.9)	N/A	0.23$\pm$0.0043	1.27$\pm$0.00946	4.85$\pm$0.0166	5.39$\pm$0.0176	7.1$\pm$0.098	0.18″
	18h28m33.35s	26d50m52.53s	N/A	24.71$\pm$0.02	22.26$\pm$0.008	19.47$\pm$0.0037	19.2$\pm$0.0035	18.36$\pm$0.015
65	277.152549(430.)	26.8653645(380.)	2.42$\pm$0.00887	3.94$\pm$0.0116	9.63$\pm$0.0276	6.41$\pm$0.0218	2.34$\pm$0.0225	4.31$\pm$0.0968	0.15″
	18h28m36.61s	26d51m55.31s	22.44$\pm$0.004	21.63$\pm$0.0032	20.06$\pm$0.0031	19.17$\pm$0.0037	20.11$\pm$0.01	18.91$\pm$0.024

Full text

††journal: ApJ

\newdateformat

monthyearday\THEYEAR \monthname[\THEMONTH] \twodigit\THEDAY

1 Introduction

The WISE mission (Wright et al., 2010) identified cool brown dwarfs (BDs) using their very red ([3.4]–[4.6] or hereafter W1-W2) colors (Kirkpatrick et al., 2011). The most extreme of these objects, with effective temperatures $T_{\rm eff}$  $<$500 K, have been typed as Y dwarfs (Cushing et al., 2011; Kirkpatrick et al., 2012). With W1-W2$>$4.2 mag, WISEP J182831.08+265037.8 (hereafter WISE 1828+2650) was identified as one of the reddest of this small group of only two dozen Y dwarfs. Follow-up photometry and spectroscopy with the Hubble Space Telescope (HST), the Spitzer Space Telescope, and various ground-based facilities found weak emission between 1.0-1.7 $\mu$m marked by absorption due to H2O, CH4 and NH3 (Cushing et al., 2011), extremely weak emission in the $K$ band (2.2 $\mu$m), followed by a sharp rise out to 5 $\mu$m.

Astrometry from Keck, Spitzer and HST determined distances to the WISE Y dwarf sample ranging from 5-15 pc (Beichman et al., 2013; Martin et al., 2018; Kirkpatrick et al., 2019). With distances serving to constrain the absolute luminosity and typical ages of 3-10 Gyr inferred from their transverse motions, it was possible to fit the photometry and spectroscopy to evolutionary models to estimate masses from 5-10 $\,M_{\rm J}$. Beichman et al. (2014) and Kirkpatrick et al. (2019) put WISE 1828+2650  at a distance of of 9.93$\pm$0.23 pc and suggested an effective temperature of 400K and a mass around 5 $\,M_{\rm J}$.

WISE 1828+2650 has proven to be exceptional even within the unusual class of Y dwarfs. Its extreme W1-W2 color, the difficulty in simultaneously fitting models to the 1-2 and 3-5 $\mu$m photometry, and WISE 1828+2650’s position at least 1 mag above the Y/T dwarf locus in the H-W2 color-magnitude diagram (Kirkpatrick et al., 2019, Figure 8) highlighted the challenges of making satisfactory models for such cold objects. Leggett et al. (2013, 2017) suggested that WISE 1828+2650  might be a binary system which would account for at least 0.75 mag of its separation from the Y/T locus. Most recently, Cushing et al. (2021) used 0.7-1.7 $\mu$m HST spectroscopy and a new generation of models (Marley et al., 2021, the Sonora Bobcat models) to bolster the idea that WISE 1828+2650  is an unresolved binary consisting of roughly equal mass objects with T_{\rm eff}$$\sim275-350 K. However, there remain challenges with the model ages ($<$1 Gyr for a single object) and the subsolar values of [M/H] $\sim-0.5$ and [C/O] $\sim-0.6$ required to fit the existing data (Cushing et al., 2021).

With the expectation that JWST would help resolve these theoretical difficulties, we set out a program of imaging and spectroscopy, particularly in the key 3-10 $\mu$m region. The JWST NIRCam imaging data permit a search for faint companions at separations and sensitivity levels not previously possible as well as anchoring the fluxes of the the higher-spectral resolution NIRSpec data. The spectroscopy provides critical diagnostics of physical conditions, composition, the presence or absence of clouds, and surface gravity in the 3-5 $\mu$m region where Y dwarfs emit most of their energy.

This paper describes NIRCam imaging in six filters spanning 1-5 $\mu$m undertaken by the NIRCam Guaranteed Time Observation (GTO) team under PID#1189. Companion papers will give results from NIRSpec and MIRI spectroscopy, and from a deep NIRISS search for a close companion. $\S$2 describes the observations and $\S$3 the data reduction procedures and results. $\S$4 addresses the search for and limits to the presence of a close companion to WISE 1828+2650 while $\S$5 presents fits to the spectral energy distribution of WISE 1828+2650 and estimates of derived physical parameters. The paper ends with a search for wide field companions ($\S$6) and concluding remarks.

2 Observations and Data Reduction

Table 1 describes the observing parameters for this program, using the long and short wavelength modules of NIRCam to observe WISE 1828+2650  simultaneously in a combination of narrow, medium and wide filters between 0.9 and 4.7 $\mu$m. Data were obtained on 2022-07-28 UTC. The NIRISS full frame imager was used to obtain a long exposure in the F480M filter in a deep search for a close companion. The NIRISS results will be presented separately, but the photometry at that wavelength is presented here for comparison with models. NIRSpec and MIRI spectroscopy were also obtained at this time and are discussed separately.

NIRCam images in the six wavelength bands and the single NIRISS band were downloaded from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute, specifically the ”…i2d.fits” files which are the result of mosaicing the four dither positions obtained at each wavelength. The specific observations analyzed can be accessed via https://archive.stsci.edu/doi/resolve/resolve.html?doi=10.17909/jxva-9x26 (catalog DOI). Figure 1 shows a three- color, full-frame NIRCam image of the field around WISE 1828+2650 and is composed of images obtained at F162M (blue), F335M (green) and F470M (red) with a close-up around the position of WISE 1828+2650. Only one object in the frame appears with the highly red color characteristic of a cool Y dwarf.

3 Analysis

The values listed here use image files created on 2022-11-11 using calibration software version 11.16.14 and photometric reference data from jwst_nircam_photom_0114.fits.

3.1 Photometry

Table 2 shows the results of standard aperture photometry using the astropy/photutils package (Astropy Collaboration et al., 2013, 2018, 2022; Bradley et al., 2020) with the image files created on 2022-11-11 and the calibration software version 11.16.14 with photometric reference data from jwst_nircam_photom_0114.fits. We selected an aperture size corresponding to 70% encircled energy and the appropriate aperture correction from the CRDS database111https://jwst-crds.stsci.edu/. The F090W filter was the weakest detection with the total flux dependent on the selected aperture due to the poorly defined Point Spread Function. For small apertures with encircled energy $<$ 50%, the flux estimates in F090W were variable at the 20-30% level. Beyond encircled energy of 50%, the flux estimates were consistent at the 10% level, shown in the total uncertainty in Table 2. We adopted a large radius of 9.25 pixels encircling 85% of the total energy for the F090W filter. We also obtained photometry using the NIRISS imager in full frame mode (Willott et al., 2022) using the same procedure as described above.

Figure 2 shows the JWST results along with previous data from Spitzer, WISE, HST and Keck (Kirkpatrick et al., 2011; Leggett et al., 2015; Kirkpatrick et al., 2019; Cushing et al., 2021). WISE 1828+2650 is detected at high signal-to-noise (SNR) in all filters with statistical uncertainties of just a few percent. However, we adopt a higher level of uncertainty for the absolute calibration in the model fitting. Successive MAST releases of these data from August-November 2022 have shown significant variations in the absolute calibration of the NIRCam filters used here. We adopt a uniform 10% uncertainty for the absolute photometric calibration in all filters.

We note the consistency between the JWST data and previous observations although different passbands can be expected to yield significant differences given the highly structured nature of the SED of cool BDs. The close agreement between multiple facilities at the wavelengths of peak emission around 4-5 $\mu$m is consistent with the general lack of variability in the brightness of WISE 1828+2650. Brooks et al. (in preparation) investigated the variability of several hundred cold BDs observed in the Kirkpatrick et al. (2021) astrometric program using Spitzer. For WISE 1828+2650 they found a limit of 3% ($3\sigma$) to the variability in IRAC Ch2 ([4.5]) in 28 observations spanning almost 8 years. A model fit based on the cloudless Sonora-Bobcat models (Marley et al., 2021) is also plotted and will be discussed in $\S$ 5.

3.2 Astrometry

As described in (Kirkpatrick et al., 2019),WISE 1828+2650 has a high proper motion ($\sim$1″/yr), mostly in right ascension, and a parallax corresponding to a distance of 9.93$\pm$0.23 pc. The values listed in Table 3 were used to establish the pointing of the NIRCam, NIRSpec and MIRI observations in this program.

The F360M image was used to obtain the position of WISE 1828+2650. The high signal to noise ratio in this filter would imply a nominal positional accuracy of FWHM/(2$\times$SNR) $<$ 5 milli-arcseconds (mas) where FWHM is the Full Width at Half Maximum of the Point Spread function. We calculated both center-of-mass and 2-D Gaussian fits to obtain the image centroid which yielded nominal accuracy of $<$0.1 pixel or $<$6 mas. Information in the FITS header was used to convert the pixel location into celestial coordinates (J2000, 2022.5699) as given in Table 3222The astrometric distortion correction in the MAST processing used the file crds://jwst_nircam_distortion_0158.asdf

A comparison of the NIRCam positions of 31 unsaturated, proper-motion corrected Gaia stars ($20\,{\rm mag}>{\rm Gmag}>$14 mag) across the 4′$\times$4′ field showed a small offset of order the measurement uncertainty with dispersion of $(\Delta\alpha,\Delta\delta)$= 10$\pm$10, 7$\pm$6) mas, demonstrating that the reference frame is accurate to $\sim$5-10 mas. However, Figure 3 shows that there remains some coherent distortions across the field of view.

The predicted position of WISE 1828+2650  was calculated at the JWST epoch incorporating stellar parallax (Smart & Green, 1977) using the rectangular coordinates of JWST’s location in the solar system as given in the FITS header (reversed to give the coordinates of the Sun). The uncertainty in the F360M position is the combination of the centroid uncertainty and the uncertainty in the reference frame derived from the Gaia stars, $\pm$2 mas, for a combined uncertainty of $\pm$4.4 mas. The difference in the predicted vs observed positions is within the uncertainties derived from a Monte Carlo run in which the Kirkpatrick et al. (2021) values were varied according to their quoted uncertainties. The Monte Carlo distribution yielded uncertainties at the JWST epoch of 6.4 mas in right ascension and declination. The Predicted-JWST position difference is consistent with the existing parallax and proper motion values for WISE 1828+2650.

4 Search for Close Companions

4.1 Near-IR Observations

Previous observations with Keck and HST revealed no evidence for multiplicity at the level of 0.1″ (Beichman et al., 2013), corresponding to orbital separations of $>$1 AU for equal brightness components. However, the Keck observations were done with the 40 mas/pixel wide-field camera mode with a typical resolution of $\sim 0.16\arcsec.$ New Keck observations were obtained 2022-Sep-08 UT with NIRC2 behind the laser guide star adaptive optics system (Wizinowich et al., 2000) in the narrow-field camera mode with a pixel scale of $0.009942\arcsec$.

Observations of WISE 1828+2650 were made in the $H$ filter $(\lambda_{o}=1.633;\Delta\lambda=0.296~{}\mu$m) with an integration time of 300 seconds per frame in a standard 3-point dither pattern that is used with NIRC2 to avoid the noisier lower-left quadrant. The transparency and seeing during the night was highly variable and not all frames detected the BD. A total of 12 frames at 300 seconds each were acquired yielding a total on-source integration time of 3600 seconds.

The science frames were flat-fielded and sky-subtracted. The flat fields were generated from a median average of dark subtracted flats taken on-sky. The flats were normalized such that the median value of the flats is unity. The sky frames were generated from the median average of the dithered science frames; each science image was then sky-subtracted and flat-fielded. The reduced science frames were combined into a single combined image using a intra-pixel interpolation that conserves flux, shifts the individual dithered frames by the appropriate fractional pixels, and median-coadds the frames. The final resolutions of the combined dithers was determined from the full-width half-maximum of the point spread functions: 7.1 pixels = 0.071″.

The sensitivity of the final combined AO image was determined by injecting simulated sources azimuthally around the primary target every $20^{\circ}$ at separations of integer multiples of the central source’s FWHM. The brightness of each injected source was scaled until standard aperture photometry detected it with $5\sigma$ significance. The resulting brightness of the injected sources relative to WISE 1828+2650 set the contrast limits at that injection location. The final $5\sigma$ limit at each separation was determined from the average of all of the determined limits at that separation, and the uncertainty on the limit was set by the rms dispersion of the azimuthal slices at a given radial distance. The Keck data have a sensitivity close-in of $\Delta H\approx 1.5$ mag at 0.071″; the final sensitivity curve for the Keck image is shown in (Figure 4). No close-in stellar companions were detected and the FWHM of the WISE 1828+2650 is consistent with the brighter nearby star to the southeast (Gaia DR3 4585337218702066560). WISE 1828+2650 is $\Delta H=3.4\pm 0.2$ mag fainter than Gaia DR3 4585337218702066560.

The $H$-band magnitude of WISE 1828+2650 as measured by the Keck data is $H=22.9\pm 0.2$ mag. Using the JWST F162M filter as a proxy for the H-band filter, this measurement is consistent with the infrared photometry determined directly from the JWST images (see Table 2).

4.2 NIRCam Observations

Although the aperture of JWST is smaller than that of Keck, JWST offers a highly stable point spread function (PSF) and much greater sensitivity at wavelengths where cold BDs are brightest. For the JWST data reported here, the best combination of spatial resolution and SNR in the JWST data comes in the F360M filter. A simple Gaussian fit of the image of WISE 1828+2650 compared to fits of two nearby stars shows no evidence for extent: FWHM(WISE 1828+2650)= $0.150\pm 0.001$″ compared with $0.153\pm 0.001$″and $0.150\pm 0.001$″for the two stars.

We explored the data further by constructing realistic models of the PSF and searching for a potential companion at all separations within 0.25”. Anderson & King (2000) and Anderson (2016) developed a novel technique applicable to multiple instruments on HST where they construct an “effective” PSF (ePSF) from bright single stars within a given image. The ePSF is intended to describe the contemporaneous realization of the theoretical PSF on the detector pixels given the wavefront during the observation. They build the ePSF first by identifying a sample of many bright stars in the image that appear to be singles and without contamination from cosmic rays or other stars. These stars are all centered at various positions within their peak pixel, giving many samplings of the distribution of flux of the center of the PSF. The initial ePSF is then generated by interpolating each PSF by the factor of user-defined oversampling, and taking the median. Then, an iterative process begins, comparing the ePSF to all stars in the sample, evaluating the median residual, and adding that on to the ePSF model. For our case, we generated an ePSF model 4x oversampled, as recommended in Anderson (2016), after 20 iterations of the ePSF calculation, over 9x9 detector pixels. This analysis was performed using publicly available PSF building tools from the python package photutils (Bradley et al., 2020).

We chose the F360M data to perform this analysis as the number of bright sources in the F470N data was small which limits the production of a reasonable ePSF model. Also, the diffraction limit in F470N is 30% larger than in F360M, limiting the sensitivity to close companions. We also avoid the F335M filter due to the low signal to noise of WISE 1828+2650. Of the four pointings, two had reliable data for WISE 1828+2650 without nearby cosmic rays or bad pixels contaminating its flux. Within each integration, we constructed a separate ePSF model from 57 and 63 separate stars (integration number 2 and 4, respectively), excluding WISE 1828+2650 from the sample.

We then made a double-PSF fitting code that takes a cutout array of the data as input (here a 9x9 pixel array centered around WISE 1828+2650) and then fits the best-fit binary model using the ePSF, an approach similarly implemented on HST (De Furio et al., 2019). We use the python module PyMultiNest (Buchner et al., 2014) that performs the Nested Sampling Monte Carlo analysis using MultiNest (Feroz et al., 2009) to derive the best-fit binary ePSF model to the data, by maximizing our chi-sqaured likelihood statistic. Our model consists of six parameters: x and y center of the primary, flux normalization of the primary, separation between the centers of the primary and secondary, the position angle of the center of the secondary relative to the primary, and the difference in magnitude between the secondary and primary. We define flat priors with -1.5 $\leq$ xcen $\leq$ 1.5, -1.5 $\leq$ ycen $\leq$ 1.5, 0.0 $<$ flux normalization $<$ 40.0, 0.01 $\leq$ separation $\leq$ 4.0 pixels, 0.0 $\leq$ position angle $<$ 360.0, and 0.01 $\leq$ difference in magnitude $\leq$ 6.0 mag.

To both images of WISE 1828+2650 in F360M in question, the code converges to the edge of the prior in separation (4 pixels) and approaching that in difference in magnitude (5.6 and 5.8 mags in each). If instead we force the code to fit within the core of the PSF (0.01 $\leq$ separation $\leq$ 2.0 pixels) and exclude wide separations, the best fit converges to 0.14${}^{+0.04}_{-0.03}$ pixels in separation and $\Delta$mag=2.44${}^{+0.34}_{-0.36}$ in Frame #2, and 0.28${}^{+0.13}_{-0.06}$ pixels in separation and $\Delta$mag=3.06${}^{+0.74}_{-0.68}$ in Frame #4 (errors are 68% confidence interval). See Figure 5 for the comparison of our binary PSF model to the data.

However, our binary PSF-fitting code will always find a best fit binary solution regardless of whether a true binary is present. If the object in question is a single point source, the code will either fit a companion to the brightest residuals in the background (e.g. the 4 pixels in separation and $\Delta$mag $\sim$ 5.5-6 mag initial fit) or to the residuals in the core of the source as models are photon-noise limited and have uncertainty (e.g. the resulting best fit when forced to fit a companion at separations $<$ 2 pixels).

In order to determine whether we can recover a companion with these values, we constructed many artificial binaries from the ePSF models at various separations and differences in magnitude with the same signal to noise as our F360M data. Then, we ran our double-PSF fitting code and a modified version that just fits a single PSF (3 variables: xcen, ycen, and flux normalization) on those same artificial binaries. PyMultiNest also calculates the evidence of the model in question by integrating over the posteriors. Trotta (2008) define a difference in the log-evidence between two models of 5 as being strong evidence for one model over another (with a probability of 0.993 that the higher evidence model is preferred). For a given companion at some separation and difference in magnitude, we compare the log-evidence of the binary-PSF model to that of the single-PSF model.

In Figure 6, we show the log-evidence difference between the binary and single ePSF model fits based on separation and difference in magnitude of our artificial binaries. We define our sensitivity as the point where the difference in log-evidence between the binary and single PSF models equals 5. With our binary fitting tool and ePSF models, we can resolve companions down to $\sim$ 0.8 F360M pixels (0.05″) in separation for an equal brightness companion and are sensitive to companions at $\Delta$mag=2.5 beyond 2 F360M pixels (0.126″). See Figure 7 for the posteriors for a fit to a binary at 1.25 pixels in separation and $\Delta$mag=1.0 and 3.5 . For the $\Delta$mag=3.5 companion, the parameters (separation, position angle, and difference in magnitude) are unconstrained.

In addition, we fit a single PSF model to the WISE 1828+2650 data set. The difference in log-evidence between the single and binary PSF models is 3.5, less than the threshold for a detection. In Fig. 5, we show both the single and binary PSF models compared to the data, demonstrating how the binary fit does not significantly improve the residuals. Therefore, we are confident that the best-fit binary model solution to WISE 1828+2650 in the F360M filter is not a true detection of a companion, and that we can rule out an equal mass companion beyond 0.5 au and a $\Delta$mag=2.5 companion beyond 1.25 au. Fainter companions at larger separations can also be ruled out although these would not help resolve the problem of excess brightness of WISE 1828+2650 relative to model predictions. Such models generally require an equal mass system as discussed below ($\S$5).

The incidence of binarity among BDs is generally low, 10%-30%, compared with higher mass stars (Burgasser et al., 2007; Raghavan et al., 2010). While less is known about the multiplicity of the coldest T/Y BDs, there is evidence that the incidence of binarity may be even lower than for warmer BDs, with Opitz et al. (2016) failing to find any companions in the range of 0.5-2 AU among the five Y dwarfs they examined with AO imaging. Similarly, Fontanive et al. (2018) estimated a companion frequency $<$ 10% for T-Y dwarfs with separations tightly peaked at 3 AU, e.g. Luhman 16 AB (Luhman, 2013). However, it should be noted that separations $\leq$ 3 AU are not well explored. If indeed WISE 1828+2650 is a binary as inferred from the spectral modeling presented by Cushing et al. (2021), then high resolution JWST NIRSpec spectra may reveal a double lined system, given an appropriate orbital plane inclination. Furthermore, two 5-10 $\,M_{\rm J}$ objects orbiting at 0.5 AU would have a period of 3-5 years and a typical orbital velocity of $\sim$5 km/s which might be discernible in multiple epochs of high SNR spectra to yield masses for the two objects.

5 The Spectral Energy Distribution and Model Fits

The broad wavelength coverage of JWST NIRCam photometry from 1 to 5µm is ideal for characterizing the spectral energy distribution (SED) and inferring the bulk atmospheric properties of BDs. We include thirteen ground- and space-based photometric points (i.e, Table 2) in our spectral energy distribution fitting. We considered a number of models for this comparison: Sonora Bobcat and Cholla (Marley et al., 2021), ATMO (Phillips et al., 2020) chemical-equilibrium (CEQ) models, ATMO chemical non-equilibrium models with strong vertical mixing (CNEQ-strong), and the Linder et al. (2019) models. We selected cloudless Sonora models333https://zenodo.org/record/1309035#.YwqSGbTMKUk,444https://zenodo.org/record/5063476#.YwqSTrTMKUk as the most up to date compared, e.g. with older COND models, or the recent Linder et al. (2019) models which are limited to low mass objects ($<$2 $\,M_{\rm J}$). The Sonora Cholla models do not go to sufficiently low temperatures for this study and the ATMO 2020 models gave similar results but with higher reduced chi-squares than the Sonora Bobcat models.

We defer discussion of the new Leggett et al. (2021) models to a companion paper on the NIRSpec low and high resolution spectroscopy (Lew et al in preparation). Leggett et al. (2021) noted that their new model failed to fit the total luminosity of WISE 1828+2650 as a single object and assumed an equal mass binary.

We used the Sonora model grid that comprises temperatures from 200 to 600 K, gravities from log(g) of 3.0 to 5.5 $\,\mathrm{m}s^{-2}$, metallicities [M/H] from -0.5 to 0.5. The ATMO model grid assumes an atmosphere with solar metallicity and span a range of temperature from 100 K to 900 K and of gravity from $10^{2.5}\,\mathrm{c}ms^{-2}$ to $10^{5.5}\,\mathrm{c}ms^{-2}$. The ATMO chemical non-equilibrium models with strong vertical mixing assume an eddy diffusion coefficient of $10^{6}\,\mathrm{c}m^{2}s^{-1}$ at a gravity of $10^{4.5}\,\mathrm{c}ms^{-2}$ and decreases with higher gravity (see Figure 1 in Phillips et al. 2020. We linearly interpolated the models to construct a model spectra grid with a smaller spacing in temperature, gravity, and metallicity. We adopt a distance of 9.93 pc reported by Kirkpatrick et al. (2019).

In the model fitting process, we examined two cases: a single object or a binary with different temperatures. For the SED fitting, we performed the least-squares fit to the thirteen photometric points. For the binary case, we require the two objects to share the same age and metallicity. To enforce the equal-age constraints, we first calculate the age based on the free parameters temperature $T_{1}$ and gravity $g_{1}$ of one component of the binary using the Sonora evolution model. We then calculate the expected gravity $g_{2}$ of the second component based on the calculated age and sampled temperature $T_{2}$ of this object. Therefore, both components of the binary share the same age, where $g_{2}$ is thus a derived value based on $T_{1},g_{1},T_{2}$ and $M$. We calculate the chi-squared values of the binary models over the grid of temperature, gravity, metallicity, and radius of the binary components. We use the bootstrapping method to estimate the uncertainties of the fitted parameters. With the bootstrapping method, we randomly re-sample the thirteen photometric points for 10,000 times and refit the models to the resampled datapoints with the least squares method. We then calculate the 99.7 percentile range of the fitted parameters as the 3$\sigma$ confidence ranges, as shown in Table 4.

In Table 4 and Figure 8, we show the best-fit parameters and display the corner plots associated with the fits to show the correlation between the parameters in our SED fitting. In the binary case, the chi-squared map suggests that the primary shares a similar temperature ($\Delta T<$50K) to the secondary when both components have non-negligible radii ($R_{1}$ and $R_{2}>0$). We also calculate the combined radius, which is the square root of the sum of the two radii squared ($\sqrt{R_{1}^{2}+R_{2}^{2}}$), for the binary case. The combined radius is moderately constrained, with a 99.7 percentile range of 1.15-2.83 $R_{J}$ and a median value of 1.67$R_{J}$, where the best-fit radius of the single model is 1.67$R_{J}$. Both the single and binary fits favor effective temperature of around 330K, low metallicity ([M/H = -0.5]), and low gravity ($\log(g)=4$). However, the best fit models suggest that the single model requires a lower mass object (2$\,M_{\rm J}$), while the binary model requires two higher mass objects ($\sim$4$\,M_{\rm J}$). Our single model has a lower reduced $\chi^{2}$ than the binary model, although both are too high to adequtely represent the data. We note that the best-fitted radius of 1.67 $R_{J}$ in the single BD case may be unphysically large, and that an equal-temperature binary may be preferred in this case, as suggested in previous studies of WISE 1828+2650 (e.g., Beichman et al., 2013; Leggett et al., 2013, 2017; Cushing et al., 2021; Leggett et al., 2021)

A striking characteristic of the SED fits for both the single and binary cases is their predicted youth, with best-fit ages between 0.3-0.7 Gyr which is at odds with the likely dynamical age of WISE 1828+2650, although these young ages are not well constrained in our model. As previously noted (Beichman et al., 2014), WISE 1828+2650 is not associated with any known young cluster. This original conclusion is bolstered by application of the Banyan $\Sigma$ tool555www.exoplanetes.umontreal.ca/banyan/banyansigma.php? (Gagné et al., 2018) which puts the probability that WISE 1828+2650 is a field object at 99% independent of its as yet unknown radial velocity. We expect WISE 1828+2650 to have an age in the 2-4 Gyr range given its tangential velocity (Beichman et al., 2013), consistent with the mean age of 2.3 Gyr for a sample of BDs in the solar neighborhood (Dupuy & Liu, 2017), contrary to our modeling of the SED.

Based on the best-fit models, we extrapolate the flux density beyond the observed wavelength regions and estimate the bolometric luminosity. The estimated bolometric luminosity has $\log(L/L_{\odot})$ of around -6.5, as listed in Table 2. The bolometric luminosity of $\log(L/L_{\odot})=$ -6.5 is similar to the expected luminosity of a $\sim$400K object at 1Gyr. Our calculations indicate that the JWST broadband photometry covers about 16% of the bolometric luminosity, while the composite photometry comprising JWST, 2MASS, WISE, and Spitzer photometry accounts for around 55% of the bolometric luminosity.

The discrepancy between models for cold BDs at long and short wavelengths is well known (Beichman et al., 2014) with models which fit well at long wavelengths failing to fit at short wavelengths and vice versa. This failure is prominent in our fits to the WISE 1828+2650 data, independent of metallicity or multiplicity. In all cases, the models, which fit reasonably at 4-5 $\mu$m, fail badly at 1 $\mu$m with the models being a factor of $\sim$2 brighter than observed at 1 $\mu$m. Another striking feature is that both the F335M and F360M data points sit significantly above the predictions suggesting that the absorption in this part of the spectrum is less than expected. Further exploration of the possible temperature-pressure profile and atmospheric chemistry, such as those models in Leggett et al. (2021), is essential to understand the atmospheric processes that drive the longstanding challenges in fitting the flux at near-IR and mid-IR wavelengths.

The SED fits to new JWST observations using two independent sets of models leave us in the uncomfortable position of either accepting a single object much younger than expected dynamically, or a near equal-mass binary system for which the NIRCam data finds no evidence at separations $>$0.5 AU. Potentially the NIRSpec’s high resolution spectroscopy will reveal a double-lined system which might resolve this problem. However, whether WISE 1828+2650 turns out to be a double, Figure 2 and the high $\chi^{2}$ values indicate significant deviations between the models and the emission of WISE 1828+2650.

At the fitted effective temperature of around 300K, it is possible that sulfide, chloride, and water clouds form near or within the photosphere and affect the emission spectra (e.g., Morley et al., 2012, 2014b, 2014a). However, Cushing et al. (2021) SED fitting results with Morley et al. (2014b) cloud models at solar metallicity give a higher reduced chi square than that with Sonora bobcat cloudless models with subsolar metallicity. Further exploring cloudy models under various atmospheric chemistry and metallicity is essential to understand the role of clouds in shaping the SED of WISE 1828+2650. It is possible that the rotation rate and observed inclination of WISE 1828+2650 cause significant luminosity deviations from a typical BD of similar temperature. Deviations as high as 20% have been estimated in comparing pole-on vs. equator-on viewed BDs at high rotation rates (Lipatov et al., 2022), but it is currently unknown if this effect occurs for WISE 1828+2650 . Important clues will come from the NIRSpec and MIRI spectroscopic observations spanning 1-12 $\mu$m at low and high resolution which are currently being analyzed.

6 A Search for Brown Dwarf Companions in the Entire Field

The large field of view and great sensitivity of NIRCam are two of the instrument’s great strengths. In addition to searching for a close companion we also looked for additional BD candidates which might or might not be associated with WISE 1828+2650. For example, Nonino et al. (2023) made a serendipitous discovery of a T dwarf (T${}_{eff}\sim$600 K) at a distance of $\sim 0.5$ kpc in a deep survey field adjacent to Abell 2744.

We used the source catalogs provided by Level-3 processing from STScI for our 6 filters. We selected F360M unconfused point sources (the ”is_extended” flag set to FALSE and the nearest neighbor distance $\geq$ 1″) with SNR$\geq$10. To avoid unreliable sources which appear at the edges of the field due to having less than the full coverage of 4 dither positions, we selected objects only within the central $2^{\prime}\times 2^{\prime}$ of the final F360M image. We selected sources in the other 5 filters (SNR$\geq$5) in the same way and used the F360M objects as seeds for band merging using a radius of 0.2″. We adopted the catalog values of aperture photometry based on 70% encircled value with pipeline-provided aperture correction. To enable a search for L,T and Y BDs we focused on filters optimized to find cold objects, requiring detections at F360M, F162M and F470N. Figure 9 shows 63 sources meeting these criteria. Sources around 0.5$>$[F360M]–[F470N]$>$0 and $0<$ [F162M]–[F470N] $<2$ mag are likely stars with spectral types as cool as M9 or galaxies (Pecaut & Mamajek, 2013)666https://www.pas.rochester.edu/~emamajek/EEM_dwarf_UBVIJHK_colors_Teff.txt.

Figure 9 shows only one object in the color space occupied by BDs, WISE 1828+2650 itself in the upper right corner of the plot. Finally, there are 4 other red objects with colors around [F162M]–[F470N] $>$2 mag (Table 5). A sixth source with [F360M]–[F470N]$\sim 1.2$ lies below the BD locus and appears to be contaminated by confusion with a nearby object. Visual inspection and measured FWHM of the four red objects shows two to be extended (Table 5). A more detailed examination of the images and comparison with the predicted WebbPSF777https://webbpsf.readthedocs.io/en/latest/ image size shows that #2 and #4 have FHWM in the F360M image of 0.23″and 0.43″, respectively, compared with other point sources in the field (0.15″in the mosaicked images) and the WebbPSF FHWM of 0.12-0.13″(Table 5). A third, #5, may be slightly extended 0.18″. The other two, #1 and #6, are point-like or only very slightly extended. Source #3 is WISE 1828+2650. Extra-galactic objects with similar colors are being found in on-going deep imaging programs and are suggestive of a new class of highly dust obscured galaxy (Hainline et al, in preparation).

Figure 9 includes colors from a large sample of BDs from Kirkpatrick et al. (2021) using H-band, IRAC C1 and Ch2 as analogs for the JWST filters. The plot also includes loci of six different Sonora BD models evaluated in the JWST filters and ranging in surface temperature, gravity, and metallicity: $300<T_{Eff}<1700$ K, log(g) of (3,4,5), and [Fe/H] of ($-$0.5,0,+0.5). The comparison between the models and the Kirkpatrick sample show the well known problem that as the BDs become cooler, the models have a progressively harder time fitting the data.

The conclusion of this search and visual inspection of the six JWST images is that there are no obvious cool BD candidates in the WISE 1828+2650 field, but that there are a significant number of galaxies.

7 Conclusion

This examination of the new 1-5$\mu$m NIRCam data for WISE 1828+2650 has confirmed what has been previously known about this source. It remains among the reddest and most challenging of the Y dwarfs to fit using existing models. Both the high SNR data at F360M and the data from the Keck telescope at 1.6 $\mu$m have failed to reveal a near-equal mass companion beyond 0.5 AU. If, as has been suggested, a binary system offers a better chance of fitting the photometric models, the companion must orbit very close to WISE 1828+2650 itself, $<$0.5 AU. Such a companion might reveal itself through double lines in the spectroscopy. However, as described above, even the binary model fails to provide an improved fit to the existing photometric data. Further insights into the properties of this enigmatic object will come with the analysis of the NIRCam and MIRI spectroscopic data forthcoming shortly in other publications.

8 Acknowledgements

We must first acknowledge the many years of effort by thousands of scientists, engineers and administrators who made JWST such a dramatic success, exceeding many of its most important requirements. This international collaboration should be an inspiration to us all. The NIRCam team at the University of Arizona and Lockheed Martin’s Advanced Technology Center brought great skill, expertise and dedication to realize the full power of this instrument. M. De Furio is grateful for support from NASA through the JWST NIRCam project though contract number NAS5-02105 (M. Rieke, University of Arizona, PI).

We wish to thank Davy Kirkpatrick for valuable discussions about WISE 1828. Some of the research described in this publication was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. M. De Furio benefited from support from JPL’s Strategic University Research Partnership (SURP). Doug Johnstone is supported by NRC Canada and by an NSERC Discovery Grant. L.A. acknowledges support by the Canadian Space Agency under contract 9F052-170914/001/MTB.

Some of the data presented herein were obtained at the W. M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W. M. Keck Foundation. The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Maunakea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain.