Backyard Worlds: Planet 9 Discoveries

• Backyard Worlds: Planet 9 is a citizen-science project that analyzes time-resolved infrared images to detect faint moving objects like brown dwarfs and potential planetary bodies.
• The project employs a hybrid discovery pipeline that combines visual flipbook inspections with automated machine-learning techniques to surpass traditional survey limits.
• Discoveries include over 3,000 motion-confirmed ultracool dwarf candidates and benchmark wide companion systems, refining our understanding of substellar evolution and outer solar system dynamics.

Searching arXiv for recent and foundational Backyard Worlds: Planet 9 papers. Found relevant arXiv papers spanning the project’s discovery pipeline, ultracool dwarfs, wide companions, spectral binaries, and Planet Nine context. Backyard Worlds: Planet 9 is a Zooniverse-powered citizen-science search of time-resolved WISE and NEOWISE infrared imaging that asks volunteers to inspect short animations composed of difference images constructed from time-resolved WISE coadds in order to find nearby brown dwarfs and potentially distant solar-system planets such as the hypothesized Planet Nine (Kuchner et al., 2017). Across its published outputs, the project has evolved from a proof of principle based on visual discovery of faint moving sources into a large discovery program for ultracool dwarfs, wide benchmark companions, rare metal-poor substellar objects, and unusual systems that standard catalog-only searches missed or did not prioritize; by 2026 it had yielded 3,006 new motion-confirmed L and T dwarf candidates, together with 28 new candidate comoving companions to higher-mass stars and 9 candidate ultracool binaries (Schneider et al., 1 Apr 2026).

1. Origins, naming, and scientific remit

The project was created because previous proper-motion searches had not fully exploited the full depth of WISE imaging, even though WISE had already proven exceptionally effective for discovering cold nearby brown dwarfs and for searching for outer solar-system objects (Kuchner et al., 2017). Its dual scientific motivation is explicit in the early description: nearby brown dwarfs can be identified by distinctive WISE colors and measurable proper motion, while distant solar-system bodies can also move detectably against the background over time. The name therefore encodes two linked aims rather than a single target.

The observational basis is the Wide-field Infrared Survey Explorer and its later NEOWISE and NEOWISE-Reactivation extensions. In the foundational project description, WISE is summarized by its four infrared bands, $W1 = 3.4\,\mu\mathrm{m}$, $W2 = 4.6\,\mu\mathrm{m}$, $W3 = 12\,\mu\mathrm{m}$, and $W4 = 22\,\mu\mathrm{m}$, with the point that time-resolved coadds built from WISE+NEOWISE data can reach $\sim 1$ magnitude fainter than single exposures (Kuchner et al., 2017). This deeper, longer-baseline regime was the intended discovery space for Backyard Worlds.

The project’s publication history shows that the scientific remit quickly broadened beyond a literal search for a ninth planet. The supplied record includes not only brown dwarfs and Y-dwarf candidates, but also a hydrogen-dominated white dwarf with a luminous compact dust disk, wide ultracool companions to stars and white dwarfs, peculiar spectral binary candidates, and metal-poor T subdwarfs (Debes et al., 2019). This suggests that the project’s central methodological contribution is the recovery of faint moving infrared sources in survey data, regardless of whether those sources are substellar benchmarks, stellar remnants with infrared excess, or solar-system candidates.

2. Discovery architecture and human visual workflow

The original Backyard Worlds interface presented volunteers with animated sequences, or “flipbooks,” of infrared difference images derived from time-resolved coadds (Kuchner et al., 2017). The sky was divided into 18,240 astrometric tiles, each $1.56^\circ \times 1.56^\circ$, and each tile was split into 64 smaller $256 \times 256$ pixel subtiles for inspection. The four displayed difference images were defined as

$$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$

$$\rm difference\ image\ 2 = epoch\ 2 - median(epoch\ 3, epoch\ 4)$$

$$\rm difference\ image\ 3 = epoch\ 3 - median(epoch\ 1, epoch\ 2)$$

$W2 = 4.6\,\mu\mathrm{m}$0

These constructions suppress static sources while emphasizing moving ones as either “movers” or “dipoles.”

The image-processing chain was deliberately tuned for by-eye recognition. The display range was limited to $W2 = 4.6\,\mu\mathrm{m}$1 to $W2 = 4.6\,\mu\mathrm{m}$2, the images were smoothed with a $W2 = 4.6\,\mu\mathrm{m}$3 tophat kernel, the smoothed background was subtracted to suppress large-scale structure and electronics noise, and an $W2 = 4.6\,\mu\mathrm{m}$4 stretch,

$W2 = 4.6\,\mu\mathrm{m}$5

was applied before combining the bands into false color with W1 as blue, W2 as red, and the mean of W1 and W2 as green (Kuchner et al., 2017). Later descriptions of the project retained the same basic logic but emphasized animated $W2 = 4.6\,\mu\mathrm{m}$6 sky cutouts from unWISE time-resolved coadds, where stationary sources self-subtract and moving cold brown dwarfs appear as orange dipoles (Marocco et al., 2024).

Candidate reporting was likewise structured around human inspection. Volunteers could use the marking tool on the flipbook, post candidates on TALK with tags such as #mover and #dipole, or submit objects through project reporting forms after additional SIMBAD and VizieR checking (Kuchner et al., 2017). In later workflows, promising candidates were vetted by professional astronomers and followed up, and the project operated alongside two additional discovery channels: photometric/proper-motion cuts and an XGBoost machine-learning classifier trained on known cold brown dwarfs (Marocco et al., 2024). The result was not a replacement of human inspection by automation, but an explicitly hybrid pipeline in which visual recognition remained central.

Several published discoveries underline the importance of this human-driven step. In WISE2150-7520AB, volunteers blinked multi-epoch unWISE/WISE images separated by about 4.5 years, noticed a faint, W2-only moving source near a bright known L dwarf, and reported it to the science team; the T8 secondary was absent from prior public catalogs and would have been difficult to recover with fully automated methods alone (Faherty et al., 2019). In the planetary-mass companion W1243, the secondary lay in the halo and along a diffraction spike of the primary, which is why it was missed by catalog-only methods and instead recovered through visual blinking in WiseView (Faherty et al., 2021).

3. Validation through ultracool-dwarf and related discoveries

The first substellar discovery associated with the project was WISEA J110125.95+540052.8, identified six days after launch and later classified from IRTF/SpeX prism spectroscopy as T5.5 $W2 = 4.6\,\mu\mathrm{m}$7 (Kuchner et al., 2017). The source has $W2 = 4.6\,\mu\mathrm{m}$8, a total proper motion of about $W2 = 4.6\,\mu\mathrm{m}$9 arcsec yr$W3 = 12\,\mu\mathrm{m}$0, and a spectroscopic distance of $W3 = 12\,\mu\mathrm{m}$1 pc. The foundational paper emphasized that this object is about 0.9 magnitudes fainter than the $W3 = 12\,\mu\mathrm{m}$2 single-exposure detection limit and almost a full magnitude fainter than the faintest discoveries in previous WISE proper-motion surveys, establishing that visual inspection of coadded difference images can recover motion below prior survey thresholds (Kuchner et al., 2017).

Subsequent follow-up formalized this validation strategy at larger scale. Spitzer imaging of 95 candidate extremely cold brown dwarfs discovered by Backyard Worlds used uniform IRAC ch1 and ch2 observations to phototype candidates and combined WISE and Spitzer astrometry to confirm motion for 75 of the discoveries; nine motion-confirmed objects had best-fit linear motions larger than $W3 = 12\,\mu\mathrm{m}$3 yr$W3 = 12\,\mu\mathrm{m}$4, and the fastest-moving discovery, WISEA J155349.96+693355.2, had total motion $W3 = 12\,\mu\mathrm{m}$5 yr$W3 = 12\,\mu\mathrm{m}$6 (Meisner et al., 2020). Using the approximate late-T/Y boundary at $W3 = 12\,\mu\mathrm{m}$7 mag, that study identified five motion-confirmed sources with colors most consistent with Y spectral types, four of them likely Y1 or later, and explicitly framed them as objects beginning to bridge the gap between the bulk of the Y-dwarf population and WISE 0855−0714 (Meisner et al., 2020).

WISEA J083011.95+283716.0 became the project’s first Y dwarf candidate. It was identified as a red, fast-moving source in AllWISE and unWISE images, then characterized with Spitzer and Hubble Space Telescope imaging (Gagliuffi et al., 2020). The key photometric constraints were $W3 = 12\,\mu\mathrm{m}$8 mag and $W3 = 12\,\mu\mathrm{m}$9 mag, with a preliminary parallax $W4 = 22\,\mu\mathrm{m}$0 mas corresponding to $W4 = 22\,\mu\mathrm{m}$1 pc. The paper interpreted these colors and limits as consistent with a $W4 = 22\,\mu\mathrm{m}$2 K source, revised to $W4 = 22\,\mu\mathrm{m}$3 K with the parallax-based distance, and described the object as a plausible bridge between the broader Y-dwarf population and the coldest known brown dwarf WISE J0855−0714 (Gagliuffi et al., 2020).

The project also demonstrated that its motion-based workflow could reveal unusual sources outside the brown-dwarf census. LSPM J0207+3331 was first flagged by a Backyard Worlds volunteer as an unusual high-proper-motion WISE source with $W4 = 22\,\mu\mathrm{m}$4, and follow-up spectroscopy established it as a DA white dwarf with $W4 = 22\,\mu\mathrm{m}$5 K and a luminous compact disk with $W4 = 22\,\mu\mathrm{m}$6 (Debes et al., 2019). The study argued that careful WISE source-position measurements showed the infrared excess to be co-moving with the white dwarf rather than background confusion noise and concluded that large-scale tidal disruption events can occur at $W4 = 22\,\mu\mathrm{m}$7 Gyr past white-dwarf formation (Debes et al., 2019).

4. Wide companions, benchmarks, and dynamical fragility

One of the clearest scientific niches of Backyard Worlds is the discovery of wide benchmark systems. WISE2150-7520AB is a widely separated $W4 = 22\,\mu\mathrm{m}$8 AU very low mass L1 + T8 co-moving system consisting of the previously known L1 primary 2MASS J21501592-7520367 and a newly discovered T8 secondary found in WISE data via the citizen-science project (Faherty et al., 2019). Spitzer photometry gave $W4 = 22\,\mu\mathrm{m}$9 mag for the secondary, FIRE prism spectra showed the pair to be spectrally normal for their types, Gaia DR2 provided $\sim 1$0 mas for the primary, and the derived physical properties were $\sim 1$1, $\sim 1$2 K, and $\sim 1$3 for the L1 primary, together with $\sim 1$4, $\sim 1$5 K, and $\sim 1$6 for the T8 secondary (Faherty et al., 2019). At $\sim 1$7 erg, the system was described as the lowest binding energy system of any pair with $\sim 1$8 not associated with a young cluster, and as the widest companion system yet observed in the field where the primary is an L dwarf or later (Faherty et al., 2019).

Ross 19B provides a complementary benchmark at much colder temperature and much wider separation. Backyard Worlds volunteers identified CWISE J021948.68+351845.3 near the nearby M dwarf Ross 19A, and follow-up astrometry plus a CoMover/BANYAN $\sim 1$9-based association analysis yielded a 100% probability that the pair is physically associated (Schneider et al., 2021). The system lies at $1.56^\circ \times 1.56^\circ$0 pc, the projected separation is about 9,900 au, Ross 19A has $1.56^\circ \times 1.56^\circ$1 and age $1.56^\circ \times 1.56^\circ$2 Gyr, and Ross 19B is estimated to be T9.5 $1.56^\circ \times 1.56^\circ$3, with $1.56^\circ \times 1.56^\circ$4 K and mass $1.56^\circ \times 1.56^\circ$5 (Schneider et al., 2021). The paper emphasized both the system’s benchmark value and its dynamical fragility, with a binding energy of roughly $1.56^\circ \times 1.56^\circ$6 to $1.56^\circ \times 1.56^\circ$7 erg.

A younger and lower-mass case is BD+60 1417B, discovered when a volunteer noticed a faint co-moving source near the K0 star BD+60 1417 in WiseView (Faherty et al., 2021). Near-infrared SpeX prism spectroscopy showed the secondary, CWISER J124332.12+600126.2, to be a very red $1.56^\circ \times 1.56^\circ$8, low-surface-gravity L6–L8$1.56^\circ \times 1.56^\circ$9 source, while the primary’s lithium, color-magnitude placement, X-ray/UV activity, and TESS rotation supported an age of 50–150 Myr. Using that age and the Gaia parallax, the companion was assigned $256 \times 256$0 K, $256 \times 256$1, and mass $256 \times 256$2, at a projected separation of 1,662 au and mass ratio $256 \times 256$3 (Faherty et al., 2021). The authors placed the system in a sparsely sampled region where the formation pathway is difficult to assess.

The same theme appears in broader samples. One study identified 89 new systems containing ultracool dwarf companions to main-sequence stars and white dwarfs, with companion spectral types ranging from M7 to T9 and host spectral types ranging from G2 to M9 (Rothermich et al., 2024). Another reported 13 new widely separated T dwarf companions to M dwarf primaries, representing a $256 \times 256$4 increase in the number of known M+T systems and spanning projected separations from a few hundred au to about 7,100 au (Marocco et al., 2024). In both cases, the systems were presented as benchmarks that begin to fill the gap between directly imaged exoplanets and stellar binaries.

5. Population growth, rare subpopulations, and classification work

Backyard Worlds increasingly functions as a population-building survey rather than only a source of isolated record objects. The largest current compilation presents 3,006 new motion-confirmed discoveries, 2,357 with L-type photometric spectral types and 649 with T-type photometric spectral types, plus an additional 80 objects considered likely L or T dwarfs based on available photometry but lacking a significant motion measurement (Schneider et al., 1 Apr 2026). The same catalog identifies 28 new comoving companions to higher-mass stars and 9 sources that are candidate binary systems made up of two ultracool dwarfs of L type or later. The authors state that, if confirmed spectroscopically, this sample would more than double the number of known L and T dwarfs (Schneider et al., 1 Apr 2026).

Companion-focused surveys show similar expansion. The 89-system study reported 21 ultracool companions to FGK stars, increasing the known population of ultracool companions to FGK stars by $256 \times 256$5, and added 72 systems with separations $256 \times 256$6 au, more than tripling the known population of such wide ultracool companions (Rothermich et al., 2024). The same paper emphasized that 12 systems have separations $256 \times 256$7 au and that the widest, CW2106+2507, is a T1 candidate at a projected separation of about 38,500 au. Quantitatively, it compared the sample to known stellar binaries and directly imaged exoplanets using mass ratio and binding energy, with approximate binding energies computed as

$256 \times 256$8

after multiplying projected separation by 1.26 to account for orbital inclination and eccentricity on average (Rothermich et al., 2024).

The project has also become a source of astrophysically unusual ultracool dwarfs. A spectroscopic survey of candidate T subdwarfs selected from Backyard Worlds identified three new metal-poor T subdwarfs with inferred $256 \times 256$9, 19 new “mild” subdwarfs with $$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$0, and three metal-rich brown dwarfs with thick-disk kinematics (Burgasser et al., 2024). That paper proposed a metallicity classification extension for T dwarfs—d/sdT, sdT, and esdT—and defined a metallicity spectral index $$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$1, with an empirical metallicity relation

$$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$2

and a scatter of about 0.29 dex (Burgasser et al., 2024). The classification work is significant because it converts BYW’s discovery stream into a structured framework for ancient, metal-poor brown dwarfs.

A related line of work concerns unresolved multiplicity and atmospheric peculiarity. Three new brown dwarf spectral binary candidates discovered through Backyard Worlds were found to be poorly fit by single near-infrared standards but better fit by composite templates with component types L7+T4, L7+T4, and L7+T7 (Bravo et al., 2023). The same analysis, however, concluded that CWISE J072708.09−360729.2 may instead be a strong variability candidate. This distinction is methodologically important: the project identifies outliers, but those outliers can trace either multiplicity or heterogeneous atmospheres, and spectroscopic template fitting alone does not always separate the two (Bravo et al., 2023).

6. Planet Nine context, misconceptions, and broader significance

The project’s title remains tied to the Planet Nine hypothesis, which proposes a new planet of roughly

$$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$3

to explain apsidal clustering of distant trans-Neptunian objects, perihelion detachment, and highly inclined or retrograde long-period orbits (Batygin et al., 2019). In that sense, Backyard Worlds belongs to a wider observational and theoretical program in which the outer solar system is searched for faint moving sources and the distant Kuiper belt is used as a dynamical diagnostic.

Recent dynamical work has extended that program by arguing that long-period, nearly planar, Neptune-crossing trans-Neptunian objects with $$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$4 AU, $$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$5, and $$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$6 AU provide a new probe of Planet Nine (Batygin et al., 2024). After bias treatment, the Planet Nine-inclusive model gave $$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$7, whereas the Planet Nine-free model gave $$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$8; with the paper’s $$\rm difference\ image\ 1 = epoch\ 1 - median(epoch\ 3, epoch\ 4)$$9 statistic, the Planet Nine-free case was rejected at roughly the $$\rm difference\ image\ 2 = epoch\ 2 - median(epoch\ 3, epoch\ 4)$$0 level (Batygin et al., 2024). At the same time, direct observational searches remain incomplete: a blind shift-and-stack search using ACT observations at 98, 150, and 229 GHz found no significant detections and eliminated roughly 17% and 9% of the parameter space for 5 and 10 Earth-mass Planet 9 models, respectively (Naess et al., 2021).

A common misconception is that Backyard Worlds has been chiefly a Planet Nine detection program. The publication record summarized here indicates otherwise. The project was indeed designed with Planet Nine in mind, but the dominant empirical return has come from the identification of nearby high-proper-motion brown dwarfs, Y-dwarf candidates, wide benchmark companions, metal-poor subdwarfs, and unusual systems that are faint, blended, catalog-incomplete, or otherwise poorly recovered by standard pipelines (Kuchner et al., 2017). This suggests that the enduring significance of Backyard Worlds lies in its demonstration that citizen-science visual inspection of time-domain infrared survey data can expose a large and heterogeneous population of astrophysically valuable moving sources, many of which sit at the boundaries of current formation, evolution, and atmospheric models.