A Novel Survey for Young Substellar Objects with the W-band Filter. II. The Coolest and Lowest Mass Members of the Serpens-South Star-forming Region

Over the last 5 yr, our team has conducted a large observing program to detect and characterize the low-mass young stellar population of nearby star-forming regions using the Wide Field IR Camera (WIRCam, Puget et al. 2004) at the Canada–France–Hawaii Telescope (CFHT) with a custom-built medium-width (6%) filter centered on a water-absorption feature at 1.45 μm (henceforth the W-band filter). In this article we report the results of observations on 2016 July 14–15 of the Serpens-South region using the W-band filter along with the broadband WIRCam J and H filters. WIRCam has a field of view of ∼20' × 20' with a sampling of 0.3 arcsecond per pixel. A single WIRCam pointing centered at R.A. = 18^h30^m03^s and decl. = −02^d01^m58^s was sufficient to cover the Serpens-South cluster. We used a 21-point dithering pattern to fill the gaps between the four detectors of WIRCam and to accurately subtract the sky background. The total integration time for J, H, and W-band filters were 1890, 1920, and 12285 s, respectively. Preprocessing of the images were performed for dark frame subtraction, flat-fielding, sky subtraction, bad pixel masking, and astrometric calibration. After combining the individual exposures into final stacked images in each filter, the source extraction and photometry were performed by using SExtractor (Bertin & Arnouts 1996). In order to avoid artifacts and false detections, we selected only those sources having a signal-to-noise ratio (S/N) > 10. Absolute photometric calibration in J and H bands was performed using data from the 2MASS point source catalog (Skrutskie et al. 2006). To calibrate our W-band photometry, we first create a catalog of available photometry for objects in our field from the WISE, SDSS, Deep Near Infrared Southern Sky Survey, 2MASS, UKIDSS, AAVSO Photometric All-Sky Survey, and USNO-B1 catalogs. For any object in our field having available near-IR and additional (shorter or longer wavelength) broadband photometry, we determine a best-fit, reddened template spectrum from the SpeX Spectral Library (Cushing et al. 2005; Rayner et al. 2009). We first scale the synthetic and observed photometry of the templates to match the observed J- and H-band magnitudes and then determine the closest matching reddened template to all photometry. We then adopt a likely W-band magnitude for the object in our field as the synthetic, scaled W-band photometry of the best-fit template. By comparing each object's likely W-band flux (based on template fitting) to the measured flux in the CFHT W-filter frame, we calculate a likely W-filter zero-point. We then use the median of these likely photometric zero-points to calculate W-band magnitudes for all of the objects in the field. The median absolute deviation of the 1033 predicted photometric zero-points for our field is 0.08 mag, and the standard deviation of the mean is a mere 0.004 mag. The final catalog after combining the photometry from J, W and H bands include photometry for 15277 sources, within an uncertainty limit of <0.1 mag.

To distinguish young brown dwarfs and planetary-mass objects from reddened background objects, which plague surveys based on broadband colors alone, we followed the approach of Allers & Liu (2020, submitted) and constructed a reddening-insensitive index (Q) by combining J-, H- and W-band photometry for survey objects:

$$\begin{eqnarray}&&Q=(J-W)+e\,\times \,(H-W)\end{eqnarray} \tag{ 1 }$$

where the constant e is expressed in terms of extinction values in each of the three bands:

$$\begin{eqnarray}&&e=({A}_{J}-{A}_{W})/({A}_{H}-{A}_{W}).\end{eqnarray} \tag{ 2 }$$

Based on synthetic photometry using an M0 spectrum from Kirkpatrick et al. (2010), we adopt e = 1.85, the appropriate value for an M0 star reddened by A_V = 10 for R_V = 3.1 (Allers & Liu 2020, submitted). Our typical contaminant is a reddened background M0 star, which should have Q ∼ 0 due to a lack of steam absorption in its spectra. As expected, the majority of objects detected in our field have a Q value of ∼0 (Figure 3). Later spectral type objects with deepening water-absorption features will have progressively more negative values of Q.

Zoom In Zoom Out Reset image size

Based on synthesized J-, H- and W-band photometry of young objects (Muench et al. 2007; Allers & Liu 2013; Allers & Liu 2020, submitted) and field dwarfs (Cushing et al. 2005), low-mass stars and brown dwarfs with spectral types later than M6 have Q values <−0.6, whereas earlier spectral type contaminants will have Q values >−0.6. The uncertainty on Q, σ_Q is estimated from propagating the errors from partial derivatives of Q as a function of J, W and H:

$$\begin{eqnarray}&&{\sigma }_{Q}=\sqrt{{\sigma }_{J}^{2}+{\left(e\cdot {\sigma }_{H}\right)}^{2}+{\left(1+e)\cdot {\sigma }_{H}\right)}^{2}}\end{eqnarray} \tag{ 3 }$$

where σ_J, σ_H, and σ_W are the uncertainties on the J-, H-, and W-band photometry respectively.

To select high-confidence candidate low-mass Serpens-South members, we considered as candidates those targets with either H < 18 mag, Q < −(0.6 + σ_Q) or H > 18 mag, Q < −(0.6 + 3  ·σ_Q). As photometric errors are significantly larger for fainter objects (as well as the pool of reddened background stars that might act as contaminants), we applied a more stringent cut for objects with H > 18 mag. Thus we obtained 69 candidate low-mass objects that satisfy the above criteria. Of these, 31 sources are brighter than 18 mag in the H band and have photometric uncertainties <0.1 mag in all three bands. Figure 2 shows the J−H versus J distribution of all the sources detected with CFHT along with the 31 low-mass candidates satisfying the above selection criteria (i.e., H < 18 mag and Q < −(0.6 + σ_Q)). Figure 3 shows the Q-value distribution of the above selected 69 sources with respect to H magnitude.

We performed follow-up spectroscopic observations of seven bright targets from the list of low-mass candidates using Astronomy Research using the Cornell Infra Red Imaging Spectrograph (ARCoIRIS) on the 4 m Blanco telescope at Cerro Tololo Inter-American Observatory (CTIO) on the nights of 2016 August 19 and 22. The spectra cover a wavelength range of 0.8–2.47 μm at a spectral resolution of R = 3500. We used a standard ABBA nodding sequence along the slit to record object and sky spectra. The log of spectroscopic observations is given in Table 1. Flat-field and argon lamps were taken immediately after each set of targets. For each target we observed a nearby A0V star for telluric correction and flux calibration.

All reductions of the data taken with ARCoIRIS were carried out with a modified version of the SPEXTOOL package version 4.1 (Vacca et al. 2003; Cushing et al. 2004; Kirkpatrick et al. 2011). The individual extracted and wavelength-calibrated spectra from a given sequence of observations, each with their own A0 standard, were then scaled to a common median flux and were median combined using XCOMBSPEC in SPEXTOOL. The combined spectra were corrected for telluric absorption and were flux-calibrated using the respective telluric standards with XTELLCOR. All calibrated sets of observations of a given object were then median combined to produce the final spectrum. Normalized, smoothed spectra, offset by a constant, for the seven brighter targets in our list are shown in Figure 4.

Zoom In Zoom Out Reset image size

We identified all matches for the seven bright targets followed up spectroscopically from the 2MASS (Cutri et al. 2003), UKIDSS (Lucas et al. 2008), Spitzer and WISE catalogs. Several of our objects are too faint to have reliable 2MASS J- and H-band photometry, thus we report only 2MASS K magnitudes here. Only K-band data is available in the UKIDSS catalog. Spitzer-IRAC four-band data is obtained from Kuhn et al. (2013). The WISE photometry at the 3.4 (W1), 4.5 (W2), 12 (W3), and 22 μm (W4) bands are obtained from the AllWISE Source Catalog (Cutri et al. 2013). Our CFHT J-, H-, and W-band photometry for the seven targets is given in Table 2 and literature near-IR and mid-IR photometry from various catalogs is presented in Table 3.

We estimate spectral types for our candidates by fitting them with ultracool young spectral standards from Luhman et al. (2017), as well as old, field spectral standards taken from the SpeX prism library, taking A_V, spectral type, and age as free parameters (by utilizing separate sets of young versus old spectral standards). Our objective is to provide a classification based on the pseudocontinuum shape rather than the individual atomic and molecular features. We calculate reduced χ² goodness-of-fit values for a grid of standards at three ages: ∼1 Myr young spectral standards, somewhat older ∼10 Myr spectral standards, and old field spectral standards. Along this grid, the spectral type ranged from M0 to L6 and we reddened templates to A_V values ranging from 0 to 30, using the reddening law from Cardelli et al. (1989). These fits are calculated within the 1.4–1.8 μm (H-band) and 1.90–2.30 μm (K-band) spectral windows, avoiding the noisy 1.8–1.9 μm range between the H and K bands. The J-band window is excluded from the fit because the S/N is usually low because of high extinction. To obtain a range of best-matching templates and corresponding spectral types and extinction, we use χ² maps, an example of which is shown in Figure 5, for SERP183038-021419. To generate these, a reduced χ² value is calculated for each grid point, and then converted to a normalized probability (taking the prior that the true spectral type and A_V lie within the ranges for these parameters that we have considered). We used the resulting maps to identify the areas of parameter space with the highest probability, and used the extent of these areas to estimate errors in spectral type and A_V. In Figure 5, regions on the map shown in darker colors delineate the best-fitting combinations of spectral type and A_V. For this particular example, the five best-fitting models have spectral types of M5.5–M6.5, ages ≤10 Myr, and A_V values of 17–19. However, models with spectral types between M4 and M7, ages ≤10 Myr, and A_V values of 17–22 provide comparably good fits, with significant degeneracy between the spectral type and A_V, hence we conservatively adopt this range of spectral types and A_V for this object.

Zoom In Zoom Out Reset image size

For late M- and L-type objects, we can distinguish between old and young objects (≤10 Myr) via the distinctive peaky H-band feature seen in low surface gravity objects and the shape of the K-band continuum (Lucas et al. 2001; Allers et al. 2007; Rice et al. 2010; Allers & Liu 2013; Liu et al. 2013; Gagné et al. 2014; Schneider et al. 2014; Esplin & Luhman 2017), however, our ARCoIRIS spectra do not have the requisite resolution to reveal other spectral indicators of low surface gravity. No clear spectral indicators distinguishing young and old M0–M4 objects are available in our spectra. Of the seven candidate low-mass objects, four are in the spectral range M4–M9, with the latest one being an M7–M9 type, and three are likely early M type (<M4) contaminants. Additionally, the four targets with spectral types > M4, SERP183044-020918, SERP182918-020245, SERP183038-021419, and SERP182917-020340 are listed as young stellar objects in the MYStIX IR-Excess Source catalog by Povich et al. (2013), SFiNCs X-ray-Infrared catalog by Getman et al. (2017), as well as in the Gould Belt Survey by Dunham et al. (2015). Evidence of excess IR emission from these various catalogs confirm the youth and likely membership of our >M4 candidate low-mass targets in Serpens-South. The remaining three sources do not appear in any of the above catalogs. While we cannot robustly determine age for <M4 objects, given the high frequency of field early M dwarfs and lack of other evidence of youth, we do not assign Serpens membership to them. They are most likely field-age foreground objects. Figures 6–9 show the normalized, dereddened spectra of four low-mass, young member candidates (see below) of Serpens-South along with their best-matching standard star spectra for three subclasses and their respective best-fit A_V values. The best-fitting spectral types and extinction ranges for our four new candidate Serpens-South members are listed in Table 4. Our spectral type classifications have an average uncertainty of ±1–3 subclasses.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Given the high extinction of these new cool dwarfs in Serpens-South, we also implemented spectral fitting using various other extinction laws in the literature such as Fitzpatrick (1999), Fitzpatrick & Massa (2007), and also a steeper power law with alpha = −2.1 (Stead & Hoare 2009), to determine if choice of extinction law affects our spectral fitting results. However, these various extinction laws did not change much our recovered spectral types or best-fit χ² values. The alpha = −2.1 law produced a <0.5 subtype shift to a later spectral type for all targets.

Our initial confirmation rate for Q-index selected objects in Serpens-South stands at four out of seven candidates, however, all three early M contaminants lie extremely close to the edge of one of the four WIRCAM array detectors; upon reexamination of the filter images, it became clear that the Q values calculated for these objects are spuriously low due to cosmetic issues at the edge of the detector. For further spectroscopic follow-up (as described in S. Dubber et al. 2020, in preparation), we have avoided following up similar spurious detections at the edge of the detector.

The spectra of young objects later than M5 show a characteristic triangular peak in the H band (Cushing et al. 2005). Of the seven dereddened spectra of candidate low-mass objects, the H-band pseudocontinuum of SERP183044-020918, SERP182917-020340, and SERP182918-020245 have a definitely peaked shape, and SERP183038-021419 shows hints of a possibly peaky H band (see Figures 6–9). This feature, where the H band is dominated by strong water-absorption bands to either side of the sharp peak located between 1.68 and 1.70 μm, was first reported by Lucas et al. (2001) for very young (∼1 Myr), cool objects in the Trapezium in Orion. The triangular shaped peak is ascribed to the effects of low-pressure and low-gravity atmospheres typical of self-gravitational collapsing objects with T_eff ≤ 2700 K and is a consistent signature of youth (e.g., Allers et al. 2007; Rice et al. 2010; Allers & Liu 2013; Liu et al. 2013; Gagné et al. 2014; Schneider et al. 2014; Esplin & Luhman 2017). The triangular shaped H-band continuum of these four objects confirms that they are young, and likely members of the Serpens-South star-forming region.

Serpens-South is known to have a high fraction of young stellar objects hosting circumstellar disks (Gutermuth et al. 2008; Dunham et al. 2015). The mid-IR photometry from Spitzer and WISE can be used to identify the presence of circumstellar disks around low-mass young objects (Luhman & Muench 2008; Luhman et al. 2008).

To determine whether the four candidate low-mass members of Serpens-South have color excesses from disks in the Spitzer and WISE photometry, we compared their dereddened [K−W3], [K−W4], [K−5.8], and [K−8.0] colors with the typical photospheric colors of young stellar objects given in Luhman et al. (2010) and Esplin & Luhman (2017). The reddening of these targets are obtained from the spectral fitting procedure described in Section 3.1. Though the presence of disks can artificially redden the stars, we ignore this factor as it is beyond the scope of this paper. In Figure 10 we compare the [K−W3] and [K−8.0] colors of the low-mass members of Taurus (Esplin et al. 2014; Luhman et al. 2017) with the new members identified in Serpens-South. The dereddened colors of the low-mass candidates have a >1–2 mag excess above their typical photospheric colors given in Esplin & Luhman (2017). Similarly, the dereddened colors in [3.6–5.8] and [3.6–8.0] are significantly redder than their corresponding photospheric sequence (Esplin & Luhman 2017) for the above four low-mass targets. The sizes of the excesses of these targets are similar to those of members of Cha I (Luhman & Muench 2008; Luhman et al. 2008) and Taurus (Esplin et al. 2014) with disks, and hence the reddened colors of four late-type Serpens-South members indicate that they are likely to host disks.

Zoom In Zoom Out Reset image size

The presence of circumstellar disks provides additional evidence of the youth and membership of those objects. Additionally, the above four targets, i.e., SERP183044-020918, SERP182918-020245, SERP183038-021419, and SERP182917-020340 are listed as young stellar objects in the SFiNCs X-ray-Infrared catalog by Getman et al. (2017) as well as in Gould Belt Survey by Dunham et al. (2015). The IR excess confirms the youth and likely membership of these candidate low-mass targets in Serpens-South.

Magnetospheric accretion in young stellar objects produces emission lines that span from UV to IR wavelengths. These emission lines can be used to obtain an estimate of the accretion luminosity, and are also signatures of the chromospheric activity of the star. In the near-IR regime, Paβ at λ = 1.2818 μm and Brγ at 2.16 μm are the major emission lines present. One of the four candidate low-mass members in our list, SERP182918-020245, exhibits Paβ and Brγ emissions (see Figure 11) which confirms its youth and association with Serpens-South. The fluxes of the Paβ and Brγ emission lines are computed by directly integrating the flux within the Gaussian curve along these lines. We used the flux-calibrated, extinction-corrected spectra for this analysis. The line luminosities of these two emission lines are computed using the relation L_line = 4πd² f_line, (Alcalá et al. 2014), where d is the distance to the target, i.e., 436 ± 9.2 pc and f_line is the extinction-corrected flux. Taking into account the range of A_V values producing acceptable spectral fits and again adopting the Cardelli et al. (1989) reddening law, we find extinction-corrected fluxes of ${1.64}_{-0.62}^{+0.52}\,\times$ 10⁻¹⁴ erg s⁻¹ cm⁻² and ${2.00}_{-0.38}^{+0.24}\,\times$ 10⁻¹⁵ erg s⁻¹ cm⁻² for the Paβ and Brγ emission lines, respectively. We calculate the accretion luminosity of SERP182918-020245 using the L_acc versus L_line relationships for the Paβ and Brγ emission lines given in Alcalá et al. (2014). The estimated $\mathrm{log}({L}_{\mathrm{acc}}/{L}_{\odot })$ values from these two lines are −1.72 ± 0.53 and −2.11 ± 0.52, respectively.

Zoom In Zoom Out Reset image size

In order to estimate the physical parameters of the candidate low-mass members identified in Serpens-South, we used the apparent magnitudes listed in Table 2. For bolometric corrections in the J and K bands, we used the values derived by Herczeg & Hillenbrand (2015) and Filippazzo et al. (2015) for young M-type objects. We adopted the latest estimate on distance to the Serpens star-forming region, i.e., 436 ± 9.2 pc (see Section 1), which corresponds to a distance modulus of 8.2 ± 0.04 mag. Using the extinction values in the V-band obtained from the best-fitting template spectral types (see Section 3.1) and adopting the reddening law from Cardelli et al. (1989) (i.e., A_J/A_V = 0.282 and A_K/A_V = 0.112), the apparent magnitudes are corrected for reddening. For the Sun, we adopted M_bol = 4.73 mag. The luminosities thus calculated from J- or K-band magnitudes for the four candidate low-mass objects are given in Table 4. The uncertainty in luminosity is estimated from the quadratic sum of fractional errors in distance, photometry, extinction, and bolometric correction. The final uncertainty in each target is dominated by the error associated with the extinction measurement, which is relatively high toward Serpens-South (see Table 4). The corresponding effective temperature for each spectral range is obtained from Herczeg & Hillenbrand (2014). Assuming an age of ∼0.5 Myr for Serpens-South (Gutermuth et al. 2008), the luminosities are then compared to the 0.5 Myr isochrone from Baraffe et al. (2015) to estimate the masses of the targets. Figure 12 shows the H–R diagram of the four low-mass members along with the isochrones and evolutionary tracks from Baraffe et al. (2015).

Zoom In Zoom Out Reset image size

Two of the four low-mass targets (specifically ID numbers 4 and 6) shown in Figure 12 appear significantly brighter than expected from evolutionary model predictions for 0.5 Myr objects. Similar trends have been observed in other star-forming regions such as Taurus, Ophiuchus, Chamaeleon, etc. (e.g., Mužić et al. 2013; Herczeg & Hillenbrand 2015; Ribas et al. 2017; Zhang et al. 2018). However, the masses and luminosities inferred for ≤1 Myr objects are very uncertain and strongly model dependent. The initial radius of a young object at its birth location depends on the initial entropy and the entropy gained and lost during the main accretion phase (Stahler 1983; Hartmann et al. 1997; Hosokawa et al. 2011). Thus, the early accretion history of young low-mass stars and brown dwarfs will affect the physical properties of objects (luminosity, radius, effective temperature etc.) in star-forming regions and young clusters (although the effect of this accretion history will become negligible by ages >10 Myr). This accretion history, however, is generally not well known for any given object, and can vary significantly from object to object in a given region. Modeling by Baraffe et al. (2017) demonstrates that accretion history produces a significant spread in luminosity on H–R diagrams for a coeval population of young low-mass stars and brown dwarfs. The location of our new candidate Serpens members above the 0.5 Myr isochrone suggests that they were born with higher entropy (and thus a larger initial radius) than predicted from standard models of evolution, which do not take into account the individualized accretion history of any given object (Baraffe et al. 2015). Binarity could also be a reason for their higher luminosity beyond the model isochrones. Nonetheless, targets 2 and 5 have effective temperatures consistent with low-mass stars of mass ∼0.1 M_⊙ whereas targets 4 and 6 have effective temperatures consistent with substellar objects with mass < 0.08–0.05 M_⊙. These are the coolest and lowest mass members of Serpens-South star-forming complex identified so far.