A Statistical Survey of Peculiar L and T Dwarfs in SDSS, 2MASS, and WISE

In P1, we reported a sample of 314 objects that passed all of our selection criteria and visual verification. We had obtained spectra of a first set of 45 of these candidates and 5 of them turned out to be false positives, i.e., their spectra were not like those of ultra-cool dwarfs. The first 40 bona fide ultra-cool dwarfs were reported in P1. We reviewed the finder charts of the 5 false positives and determined that these objects indeed did not look like the bona fide ultra-cool dwarfs (e.g., were more diffuse than point-like). After refining our visual verification, as informed by our re-analysis of the 5 false positives, our total candidate L and T dwarf list was cut to 156 objects including 104 new candidates, the 40 candidates reported in P1, the original 5 false positives, and 7 new false positives. For this paper we will focus on the remaining 104 candidates that we verified to be ultra-cool dwarfs via spectroscopy and only briefly discuss the 12 total false positives in Section 4. We also recovered 276 ultra-cool dwarfs that were previously known. We discuss these in Section 5, where we analyze the statistics of the whole survey.

The priority of our survey was to identify peculiarly red ultra-cool dwarfs in the cross section of the SDSS, 2MASS, and WISE catalogs. We designed an appropriate selection criterion from the sample of L and T dwarfs in the SpeX Prism Archive. We formed synthetic photometry from their spectra by convolving with the 2MASS filter transmission profiles and integrating over the filter bandpasses. We identified objects that had $J-{K}_{s}$ colors that were >2σ redder than the median for the spectral type (red symbols in Figure 1(a)). The medians and standard deviations of the $J-{K}_{s}$ colors were taken from Faherty et al. (2009; M7–M9 and T0–T8) and Faherty et al. (2013; L0–L9). All of the red color outliers in the SpeX Prism Archive lie above the $z-J=-0.75(J-{K}_{s})+3.8$ mag line in Figure 1(a). We then applied this criterion to our 156 ultra-cool dwarf candidates and ended up with a sample of 88 peculiarly red candidates, 22 of which were already reported in P1. The synthetic colors of all of the candidates are presented in Figure 1(b), with different plotting colors used to represent objects with various spectroscopic peculiarities (discussed in Section 4). Our selection criteria were based on the photometric colors of the candidates so there are a number of objects whose synthetic colors do not appear to pass the initial color-selection criteria (discussed more in Section 4).

Zoom In Zoom Out Reset image size

In addition to selecting unusually red objects, our prioritization criterion from Section 2.2 also efficiently identified candidated unresolved L+T binary brown dwarfs. Figure 1(b) shows that they also stand out from the locus of objects on a z − J versus $J-{K}_{s}$ diagram. Late-L and early-T dwarfs are similar in brightness in the J-band but are fainter in the z- and K-bands, resulting in moderate or blue $J-{K}_{s}$ colors but red z − J colors. To fully include all potential unresolved L+T binaries, we created a second independent criterion to select these. Any object that satisfied the criterion $z\,-J\gt 0.95(J-{K}_{s})+1.45$ mag was either a candidate L+T binary or a candidate T dwarf, as the latter also have the same red-optical and near-infrared colors. This criterion selected 13 objects that were candidate binary or T dwarfs. Twenty objects satisfied both selection criteria, i.e., they were red in z − J but moderate in $J-{K}_{s}$ (top center of Figure 1(b)).

We obtained the majority of our follow-up observations (91 of 104) with the SpeX spectrograph on the IRTF in prism mode (0.75–2.5 μm; R ∼ 75–150), between 2014 October and 2016 April. The observing sequences and instrument settings were the same as those in P1. Table 1 gives observation epochs and SpeX instrument settings for each science target. We reduced all the SpeX data in the same way as in P1. Figure 2 shows all reduced spectra in order of increasing spectral type (see Section 4) and within each spectral type in order of increasing RA.

Zoom In Zoom Out Reset image size

View extended figure (5 panels)

We followed-up the remaining 13 objects in our candidate list using GNIRS on Gemini North (0.9–2.5 μm). We observed these objects in queue mode between 2015 October and 2017 May. We took the observations in cross-dispersed mode with the short-blue camera with 32 l/mm grating and a $1\buildrel{\prime\prime}\over{.} 0\times 7\buildrel{\prime\prime}\over{.} 0$ slit, resulting in a resolution of R ∼ 500. We used a standard A-B-B-A nodding sequence along the slit to record object and sky spectra. Individual exposure times were 120 s per pointing. Standard stars were used for flux calibration and telluric correction. Flat-field and argon lamps were taken immediately after each set of target and standard star observations for use in instrumental calibrations. Table 2 gives Gemini/GNIRS observation epochs for each science target.

We reduced the cross-dispersed spectra by straightening the traces, rectifying them to the vertical, and then wavelength-calibrating before extracting. We extracted the spectra using the variance weighted sum of the flux within the aperture, with the aperture radius equal to the PSF radius (usually ∼4 pixels = $0\buildrel{\prime\prime}\over{.} 60$). We modeled a local background using a linear fit to a specified background region (usually ∼8 pixels = $1\buildrel{\prime\prime}\over{.} 2$ wide on either side of the PSF ∼2 pixels = $0\buildrel{\prime\prime}\over{.} 30$ away from the PSF) and subtracted it from the spectra, which we subsequently extracted. Each set of extracted spectra was median-combined, corrected for telluric absorption, and flux-calibrated with the associated A0 calibration star. We median-combined all calibrated sets of observing sequences to produce a final spectrum. The reduced spectra were smoothed, using the IDL interpolation algorithm with a least-squares quadratic fit, to the same resolution as the SpeX standards for comparison in Section 4. The reduced Gemini/GNIRs spectra are included in Figure 2, where they are shown prior to smoothing.

We classify seven objects as unusual based on their spectroscopic peculiarities. Our assessment of peculiarity is determined by high spectral similarity to objects that have previously been classified as peculiar. The peculiar characteristics of our objects can be produced by youth, large amounts of atmospheric dust, or low-metallicity. The most interesting objects from this portion of the survey are the young, planetary-mass L7 dwarf 2MASS J00440332+0228112 (Section 4.1.2), the very red L6 dwarf 2MASS J03530419+0418193 (Section 4.1.4), and the candidate young L1 dwarf 2MASS J00133470+1109403 (Section 4.1.1).

Young ultra-cool dwarfs have low surface gravity, hence the line strengths of the gravity-sensitive features will differ from those in older objects (e.g., Lucas et al. 2001; Gorlova et al. 2003; McGovern et al. 2004; Allers et al. 2007; Lodieu et al. 2008; Rice et al. 2010; Allers & Liu 2013). The Na i (1.138 and 1.141 μm) and K i (1.169 and 1.178 μm, 1.244, and 1.253 μm) doublets are weaker because of decreased pressure broadening. The FeH features (bandheads at 0.990 and 1.194 μm) are weaker because of the decreased opacity of the refractory species. Collision-induced absorption from molecular hydrogen also changes as a function of gravity, with lower collision rates in low-gravity objects imparting a triangular shape to the H-band.

As discussed in P1, the indices developed by Allers & Liu (2013), Canty et al. (2013), and Schneider et al. (2014) have the potential to offer a quantitative gravity classification. However, our spectra have significantly lower spectral resolution, so the index measurements are more uncertain than in those studies. In addition, most of the indices do not extend into the late-L dwarfs, and thus are inadequate for classifying some of our most interesting objects. Therefore, we do not adopt spectral indices as a default gravity classification scheme. However, we do check for consistency with applicable spectral indices whenever we find peculiarities in the spectra of our L and T candidates.

We note that unusually red objects that lack some signatures of youth can still exhibit some of the spectroscopic characteristics of young objects, particularly weaker FeH bands and a triangular H-band continuum (e.g., Looper et al. 2008; Kirkpatrick et al. 2010). In these objects, such features have been attributed to high atmospheric dust content or to thicker clouds rather than to low-gravity (Looper et al. 2008; Allers & Liu 2013). In the cases of unusually red objects, we rely on the strength of the gravity-sensitive alkali (K i and Na i) lines to distinguish between young objects and field-age ($\gtrsim 0.5$ Gyr) objects with unusually high dust content.

For objects with blue near-infrared colors, we seek to determine whether the blue colors may result from low cloud opacity, low-metallicity, or unresolved binarity. In order to check the possibility of them being binaries, we consider the CH₄ in the K- and H-bands. If there is a higher abundance of CH₄ in the H-band relative to the K-band, the object is most likely a binary, as the onset of methane absorption in cooler, older objects is apparent in the K-band before the H-band (e.g., Cruz et al. 2004; Burgasser et al. 2010; Bardalez Gagliuffi et al. 2014). If we determine that the objects are not binaries, then we can also check whether or not they are metal-poor by analyzing the FeH feature in the Y- and J-bands. Although an object may be metal-poor, it may have a stronger FeH feature due to the larger relative amounts of hydrogen present and the decreased absorption by oxides such as TiO and VO (e.g., Kirkpatrick et al. 2010). If, however, an object's blue colors come from a low cloud opacity, the overall dust continuum will simply be much weaker, leaving CO and CH₄ as the dominant opacity sources (Kirkpatrick 2005).

We discuss individual objects and their defining characteristics in the next sections.

4.1.1. 2MASS J00133470+1109403 (L1)

This object is a young L1 dwarf. Compared to a normal L1.5 dwarf, 2M J0013+1109 has much weaker FeH and K i absorption lines and a triangular H-band (Figure 4). The Allers & Liu (2013) spectral indices say that this object is an INT-G object. This object is very similar to the L1 β brown dwarf 2MASS J01174748–3403258 (Burgasser et al. 2008), however, it has even weaker K i absorption lines and a redder continuum. Based on these characteristics, 2M 0013+1109 likely has lower gravity than 2M 0117–3403 and could potentially be a ∼10 Myr-old free-floating planetary-mass object.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

However, according to BANYAN II (Malo et al. 2013; Gagné et al. 2014) and the Convergent Point tool (Rodriguez et al. 2013), 2M 0013+1109 does not have a likelihood of being a part of any of the young associations and moving groups used in those works. Instead, BANYAN II calculates a 98% probability of being part of the young field when using a young prior, and 60% probability of being part of the young field when no priors are set.

4.1.2. 2MASS J00440332+0228112 (L7)

4.1.3. 2MASS J03302948+3910242 (L7)

4.1.4. 2MASS J03530419+0418193 (L6)

4.1.5. 2MASS J10265851+2515262 (L5)

4.1.6. 2MASS J10524963+1858151 (L5)

4.1.7. 2MASS J14313545−0313117 (L6)

Three of the objects show peculiarities that do not readily match those found in other individual objects. Instead, they more closely resemble combination spectra of L and T dwarfs. Burgasser (2007) and Burgasser et al. (2010) developed a technique that enables one to infer the spectral types of the individual components of a candidate unresolved binary by a goodness-of-fit comparison to a library of spectral template combinations. We adopt this technique by creating combination templates from the set of single L and T dwarfs from the SpeX Prism Library. In P1 we created templates using only the near-infrared standard objects. We have now expanded our templates to include objects from the entire SpeX Prism Library, as done in Bardalez Gagliuffi et al. (2014).

We constructed our composite template spectra in the same way as P1 but we have now used the updated absolute spectral-type dependent magnitude polynomials given in Table 10 of Filippazzo et al. (2015). We classified an object as a likely spectral type composite—a potential unresolved binary—if the ${\chi }^{2}$ (calculated over the entire 0.8–2.5 μm region, minus the water absorption bands) of the dual-template spectral fit is significantly lower than the ${\chi }^{2}$ of the single-template fit.

In addition to template fitting, where applicable, we analyzed the spectral indices defined specifically for SpeX prism spectra in Burgasser et al. (2010; for L+T binaries) and Bardalez Gagliuffi et al. (2014; for M+T and L+T binaries).

4.2.1. 2MASS J00132229–1143006 (T3.5+T4.5?)

While this object has a J-band spectrum consistent with that of a T3 dwarf (Figure 5), it is better fit by a composite template of a T3.5 + T4.5 dwarf. However, this object could instead be a blue T3 dwarf, as all the features are similar to those of the T3 dwarf except the blue color. A third explanation is that this object is a variable T dwarf that displays two distinct temperature components. An example is 2MASS J21392676+0220226, which was originally thought to be a candidate L8.5+T3.5 unresolved binary (Burgasser et al. 2010) but was later identified as a high-amplitude variable (Radigan et al. 2012). This may also be the case with 2M J0013−1143. This object only satisfies two of the six binary index selection criteria from Table 5 of Burgasser et al. (2010). Overall, because of the much better binary-template fit, we treat 2M J0013–1143 as a likely unresolved binary but note that it can instead be a blue T3 dwarf.

Zoom In Zoom Out Reset image size

4.2.2. 2MASS J00452972+4237438 (M8+L7?)

4.2.3. 2MASS J09240328+3653444 (L2+T3)

We classify 80 of our candidates as normal L dwarfs, i.e., they do not have any readily apparent peculiarities based on their comparison to SpeX spectral standards. We also identified 14 candidates as normal M dwarfs. These were included in our program likely because the i − z and z − J colors of late-M dwarfs are close to the limits of our color-selection criteria (Section 2), and because they may have been subject to flux-overestimation bias at the J-band (Section 3.3 of P1). The spectra of all these objects are included in Figure 2.

We identified five false positives in P1 and seven in the concluding portion of the survey presented here. The spectra of these objects were taken with SpeX, GNIRS, and FIRE (see P1 for instrument set up and extraction details for the Magellan/FIRE data). The spectra are presented in Figure 6 and are grouped by spectral similarity. Their synthetic colors and photometric magnitudes are presented in Tables 3 and 4, respectively. As can be seen in Figure 6, the spectra of these objects do not appear to be those of ultra-cool dwarfs. The first two of these objects in Figure 6 look like they might be warmer stars and were most likely nearby stars that were targeted by mistake. We treat these objects as false positives for the purpose of the survey statistics even though the candidates may indeed be ultra-cool dwarfs. All of the other objects have steep red slopes in the z- and J-bands, peak in the H-band, and are either relatively flat or taper off into the K-band. The z − J and $J-{K}_{s}$ colors of these objects are similar to those of ultra-cool dwarfs and would have passed the selection criteria quite easily. We also note a large discrepancy between the photometric and synthetic colors for these red objects (Tables 3 and 4). This suggests that the photometric magnitudes for these objects are not accurate. We believe that all of the latter objects are extragalactic in origin.

Zoom In Zoom Out Reset image size

We had a false positive rate of 11% from the first portion of our survey (5 unknown objects out of 45 total candidates). We refined our visual selection for the second portion based on these false positives and reduced our rate to 6.2% (7 unknown objects out of 113 total remaining candidates). We exclude these objects when discussing the statistics of our full survey and only consider the 144 confirmed new and 276 previously known ultra-cool dwarfs.

From the total 420 objects that our color criteria selected and that we confirmed to be ultra-cool dwarfs, 17 (4.0%) are peculiarly red, 7 (1.7%) are peculiarly blue, and 30 (7.1%) are candidate binaries. The number of peculiarly red objects in our sample is statistically equivalent with that of the 4.6% of red objects found in an unbiased sample of L and T dwarfs from Faherty et al. (2009). Our peculiarly red selection criterion, however, identified 9 new red objects among the sample of 92 candidates (9.8%). None are peculiarly red out of the 31 general ultra-cool dwarf candidate sample, so our selection technique successfully identified all of the peculiarly red objects in the sample. Including the previously known ultra-cool dwarfs, 6.2% of objects were red among the sample of objects that satisfied the peculiarly red criterion (17 of 274).

We had a better success rate for objects that satisfied the candidate binary criterion. Of the 51 new binary candidates, 10 (20%) were either peculiarly blue (2), or are potential binaries (8). Including the previously known ultra-cool dwarfs, 13% were unusually blue (3) or candidate binaries (20) among the sample of objects that satisfied the candidate binary criterion (173).

When we compare our newly discovered peculiarly red brown dwarfs (9) to the number of red L and T dwarfs that were already known and were recovered with our selection criterion (8), we see that we increased the sample in this color space by a factor of two. Similarly, we see that we increased the sample of candidate binaries by a factor of 1.4. With only two new unusually blue discoveries, we did not significantly impact the statistics of these types of objects. This is unsurprising, as we were not targeting these objects.

We also did not uncover many L+L or T+T binaries. Binaries where both components have similar spectral morphologies are harder to discern from single low to moderate resolution spectra, and thus are more difficult to identify. An example of this is the planetary-mass object, 2MASS J11193254–1137466, which we discovered in the first portion of the survey (L7; P1) that was later resolved into a binary system (L7+L7; Best et al. 2017a). We identified this object as peculiar based on its extremely red near-infrared colors and weak alkali absorption features but did not suspect unresolved binarity from our spectra.

Among the 17 unusually red objects from the whole survey of 420 ultra-cool dwarfs, 9 (2.1%) are young and 8 (1.9%) are red, with no signatures of youth. The ∼2% fraction of low-gravity ultra-cool L dwarfs is consistent with a ≲200 Myr age for these objects under the assumption of a constant star formation history for the Milky Way (e.g., Burrows et al. 1993; Marley et al. 1996). This is also consistent with evolutionary models such as those in Burrows et al. (2001) and Baraffe et al. (2015) that say the radii of ultra-cool dwarfs become constant after ∼200 Myr and studies of objects in young stellar associations such as Allers & Liu (2013) and Liu et al. (2016), which show that associations older than ∼200 Myr have a much smaller fraction of objects that show signatures of youth than younger associations.

For the 1.9% of objects that are unusually red with no signatures of youth, there has thus far not been a satisfactory explanation for their redness. We suggest that these objects may have reached a point where their surface gravities are not low enough to have distinguishing alkali line strengths from field objects in low-resolution spectra, but are still young enough that their dust has not completely settled. This dust could be the sub-micron particles that potentially play a significant role in the reddening of ultra-cool dwarf spectra and settle less efficiently than the 1–100 μm grains (Hiranaka et al. 2016). If this is the case, then the peculiarly red color that is detectable in their optical to near-infrared SEDs is a better indicator of moderate youth than individual absorption features. Another possible explanation is that these objects were formed in environments with higher metallicity. If this is the case, then ∼2% of objects within the sensitivity limit of the SDSS, 2MASS, and WISE surveys were born in these conditions.

Kirkpatrick et al. (2010) offer further speculation on the nature of these unusually red objects. They find that the space velocities of these objects are similar to those of unusually blue L dwarfs that do not have any signatures of low-metallicity. They suggest that these two kinds of objects could be related and their spectral appearance could be the result of different viewing angles. If clouds are not homogeneously distributed in latitude or if cloud properties such as grain size and thickness vary in latitude, then viewing an object pole-on versus equator-on would change the spectral morphology. However, if this were the case, then we would expect the number of objects with these properties to be higher, because such conditions would be ubiquitous for all brown dwarfs.

Our binary selection criterion was designed to identify L+T binaries. In these objects, both components contribute equally to the flux in the J-band but unequally in the z- and K-bands, making the z − J colors red and the $J-{K}_{s}$ colors blue. We identified and recovered 30 objects that are candidate binaries out of the 420 objects in the full survey, 17 of which are binaries with 1 L and 1 T component. We did not find any new L+L binary candidates and only two new T+T binary candidates, likely because they are difficult to identify even from their spectra, as the two components would be more similar in spectral morphology than the components in an L+T binary. These objects would look more like single objects and could only be identified as binaries through other means.

Our survey also recovered 51 known single T dwarfs and 9 candidate T+T binaries, 2 of which were new. These represent the entire population of T dwarfs in SDSS. The vast majority were already known from previous searches for T dwarfs in SDSS. We categorized both of the newly discovered objects we as candidate binaries, although they could also be highly variable T dwarfs.

As done in P1, we investigated whether the $J-{K}_{s}$ color outliers also have unusual colors in the W1 and W2 WISE bands. We have compared the objects that we discovered in this survey to known ultra-cool dwarfs in the AllWISE catalog published in Kirkpatrick et al. (2011). We determined the red and blue outliers from this data set using the median $J-{K}_{s}$ colors and standard deviations for each spectral type from Faherty et al. (2009) and Faherty et al. (2013). We confirm the results of P1 that L dwarfs with the very reddest $J-{K}_{s}$ colors are clearly distinguishable from the locus of L dwarfs on a $J-{K}_{s}$ versus $H-W2$ and $J-{K}_{s}$ versus $W1-W2$ diagram (Figure 8) mainly because of their red $J-{K}_{s}$ colors. T dwarfs with peculiarly red $J-{K}_{s}$ colors are slightly redder in both $H-W2$ and $W1-W2$, and the peculiarly blue L or T dwarfs and candidate binaries are not distinguishable from the normal population.

Zoom In Zoom Out Reset image size