ABYSS. I. Targeting Strategy for the APOGEE and BOSS Young Star Survey in SDSS-V

Historically, the YSOs that have been easiest to identify are those that have large infrared excess due to the presence of a protoplanetary disk, particularly in the mid-IR regime. In the last few decades, telescopes such as Spitzer and Wide-field Infrared Survey Explorer (WISE) have particularly expanded the census of dusty YSOs. In particular, WISE, due to being an all-sky survey, is particularly informative for targeting. Several studies have used WISE to search for YSOs (e.g., Koenig & Leisawitz 2014; Marton et al. 2016; Kang et al. 2017), but they either focused only on a specific star-forming region or had a large degree of contamination across the entire sky.

As at this stage in the survey, the goal is to create a census of sources that should be targeted for follow-up observations (rather than explicit classification of YSOs into evolutionary stages), thus we use simple color cuts in WISE photometry for this carton. To minimize a selection of very distant, highly extincted field stars (which are the main source of contamination), we impose a parallax cut as well—this implicitly requires all of the identified sources to be bright enough in the optical regime to be detected and have reliable astrometry with Gaia. We select sources satisfying the following:

• 1.  
  W1 − W2 > 0.25 mag;
• 2.  
  W2 − W3 > 0.5 mag;
• 3.  
  W3 − W4 > 1.5 mag;
• 4.  
  π > 0.3 mas.

These cuts have been evaluated against known dusty YSOs in the Orion molecular clouds (Megeath et al. 2012), and they are shown in Figure 2.

Zoom In Zoom Out Reset image size

ALLWISE photometry was used for the selection. Although there have been recent rereductions, such as unWISE or neoWISE, their improvements are primarily in W1- and W2-band photometry; W3 and W4 mostly remain as is. Longer wavelength bands lack the sensitivity of shorter wavelength bands; furthermore, they have not been observed for as long due to WISE running out of cryogenic coolant needed to suppress the telescope emission at these wavelengths. Nonetheless, W3 and W4 bands are critical for reliably identifying dusty disk-bearing stars. As such, by requiring these bands and using merged photometry, any improvements in W1 or W2 have negligible effect on the selection.

In the V0 version of the targeting, these cuts formed the basis of YSO_S1 carton. In version V0.5, the carton was renamed and split into YSO_Disk_APOGEE and YSO_Disk_BOSS, containing 28,832 and 37,478 stars respectively. In version V1, Gaia Data Release 2 (DR2) astrometry was upgraded to EDR3.

Some disk-bearing sources are too faint to have reliable Gaia parallaxes. This is usually the case for class I protostars that are still embedded in their natal envelopes, class II YSOs that have edge-on disks, or the sources that are more distant and thus have more extinction along the line of sight. Without a distance estimate, it can be difficult to separate bona fide YSOs from distant and heavily extincted red giants. As such, more stringent color cuts, not just on ALLWISE photometry but also Two Micron All Sky Survey (2MASS), are required:

• 1.  
  G > 18.5 mag, or undetected;
• 2.  
  J − H > 1 mag, H − K > 0.5 mag;
• 3.  
  W1 − W2 > 0.5 mag, W2 − W3 > 1 mag, W3 − W4 >1.5 mag;
• 4.  
  W3 − W4 > 0.8 × (W1 − W2) + 1.1 mag.

These cuts are shown in Figure 2.

In V0 version of targeting, the carton was referred to as YSO_S2; in V0.5 it was renamed as YSO_Embedded_APOGEE, containing 11,086 stars. Following the first year of operations, the carton was re-examined; approximately half of the observed sources were red giants, as shown in Figure 2 (bottom right panel). Therefore in V1 we added an additional cut

• 1.  
  H − K > 0.65 × (J − H) − 0.25 mag

to minimize the contamination by evolved stars and further restrict the selection to the parameter space that is most commonly inhabited by spectroscopically confirmed YSOs, limiting the sample to 5455 stars.

The selection criteria of disk-bearing and embedded sources rely on WISE bands W3 and W4. W3 and W4 however become less effective in regions of high nebulosity, as they saturate, producing gaps in the coverage. To fill them, we used shorter wavelength data, using a set of criteria that is tuned to autonomously find such nebulous regions:

• 1.  
  If W4 is not reported, W2 − W3 > 4 mag.
• 2.  
  If W3 and W4 are not reported, J − H > 1.1 mag.

This preferentially selects sources found in gaps of the previous cartons in discrete regions on the sky, such as, e.g., in the center of the Orion Nebula (Figure 3).

Zoom In Zoom Out Reset image size

Additionally, some of the sources that are selected by these cuts are found off of the Galactic plane and/or away from known star-forming regions. This creates an excess of targets in a narrow line following the scanning law of WISE telescope; and so, these sources appear to be suspect. Thus, we also required b < 5° and a combination of b > − 5° or l > 180°, to exclude this contamination.

This carton was introduced in V0 as YSO_S2.5. In V0.5 it was renamed as YSO_Nebula_APOGEE, containing 1112 stars.

The inner Galaxy, including the central molecular zone (CMZ), has been surveyed with Spitzer as a part of GLIMPSE and MIPSGAL programs (Carey et al. 2009; Churchwell et al. 2009; Gutermuth & Heyer 2015). These data offer a substantial improvement on sensitivity and resolution in comparison to WISE; therefore they are advantageous in targeting stars outside of the solar neighborhood.

Using the properties of massive YSOs toward the Galactic center identified by An et al. (2011), we select a sample of candidates with the following:

• 1.  
  [8.0] − [24] > 2.5 mag, i.e., very red sources, using photometry from Gutermuth & Heyer (2015);
• 2.  
  π < 0.2 mas or not detected and/or measured, to ensure the sources are distant.

In V0 of targeting, YSO_CMZ carton also imposed a spatial limit of 358 < l < 2° and −1 < b < 1° to focus solely on the CMZ. In V0.5, the carton was renamed to YSO_CMZ_APOGEE and removed the spatial restriction, allowing the sources from the entire MIPSGAL footprint. This enables to identify YSO candidates across the inner Galaxy, containing 13,170 stars.

On average, YSOs tend to be more variable than main-sequence stars (e.g., Kounkel et al. 2022b). Part of the reason for this variability is the presence of stronger magnetic fields that leads to more prominent star spots with a larger filling factor. Also the presence of protoplanetary disks that would occult the photosphere, and accretion events can lead to an increase in brightness.

To estimate variability V_x , we use multiepoch photometry by Gaia. Following Belokurov et al. (2017), we define the photometric variability in a given filter x as follows:

$$\begin{eqnarray*}&&{V}_{x}=\sqrt{\mathrm{phot}\_{\rm{x}}\_{\rm{n}}\_\mathrm{obs}}/\mathrm{phot}\_{\rm{x}}\_\mathrm{mean}\_\mathrm{flux}\_\mathrm{over}\_\mathrm{error},\end{eqnarray*}$$

where x corresponds to the Gaia G, G_BP, or G_RP bands, phot_x_n_obs is the number of observations that contributed to the photometry in a given band, and phot_x_mean_flux_over_error is the mean flux in a given band divided by its error. For strongly variable sources, the photometric uncertainty is comparable to the amplitude of variability.

Using the list of members of the Orion Complex from Kounkel et al. (2020) as a representative sample of young stars, we develop a set of criteria based on variability that allows to most cleanly preserve a large fraction of members while rejecting field stars within the volume of space surrounding Orion. This set of criteria was later applied to the entire sky and further modified to preserve the morphology of nearby star-forming regions and minimize contamination, resulting in the following:

• 1.  
  V_G > 0.02, V_BP > 0.02, V_RP > 0.02; the reference YSOs tend to be more variable than the field stars.
• 2.  
  ${V}_{G}^{0.75}\lt {V}_{\mathrm{BP}}\lt {V}_{G}$, $0.75{V}_{G}\lt {V}_{\mathrm{RP}}\lt {V}_{G}^{0.95}$; correlation of variability in different bandpasses on the order of unity appears to be a strong indicator of YSOs. On the other hand, while YSOs with different correlation in variability do exist, it is difficult to reliably separate them from the field stars, and so, many young variable stars may be excluded from the selection. The slopes of these power laws were determined from examining Figure 4.
• 3.  
  G_BP − G_RP > 1.3; hot stars do not have convective atmospheres, and thus they generally do not have spotted photospheres. While this specific cut is redder than that of the convective limit, it also minimizes the extreme contamination from the red giants. Variability among hotter stars may often be an indicator of other processes, such as, e.g., eclipsing binaries, which is not an indicator of youth.
• 4.  
  ${M}_{\mathrm{BP}}\gt 5{\mathrm{log}}_{10}{V}_{\mathrm{BP}}+11$ preferentially excludes the evolved subgiants, despite some overlap in the parameter space with bona fide YSOs.
• 5.  
  2.5(G_BP − G_RP) − 1 < M_G < 2.5(G_BP − G_RP) + 2.5; previous criteria preferentially select YSOs, but it still includes some strongly variable main-sequence stars. This selection confines the parameter space on the H-R diagram to minimize contamination.
• 6.  
  π > 0.3 mas, G < 18.5, H < 13, limiting the selection to the sources to brighter stars with reliable parallaxes. Note that the H-band cut is applied both to APOGEE and to BOSS samples, as sources with H > 13 and G_RP < 15.5 preferentially trace out more distant stars that seem to be more strongly contaminated than the sources within 1 kpc. As such, the BOSS variable sample is a subset of the APOGEE variable sample, with the faint limit in place.

These cuts are shown in Figure 4.

Zoom In Zoom Out Reset image size

The carton based on this selection has been introduced in V0 as YSO_S3. In V0.5 it has been renamed and split into YSO_Variable_APOGEE and YSO_Variable_BOSS, containing 52,691 and 47,758 stars respectively. In V1, the selection has been upgraded from Gaia DR2 data onto Gaia EDR3.

We note that the third Gaia data release (Gaia DR3) has produced a catalog of young star candidates based on their variability (Marton et al. 2022). The selection presented here was originally derived prior to the availability of these data. Out of 79,375 sources presented in Gaia DR3, our selection has 3963 stars in common with this carton, and 16,474 stars across all of the cartons. Of the remaining 20,225 candidates in Gaia DR3 that would meet our faint limit, as much as a half appear to be contamination from highly reddened distant main-sequence and red giant stars, but in future versions of the targeting definition, with some refinement, it may be possible to take advantage of this catalog. On the other hand, our current selection does not extend to as faint magnitudes (and, indeed, applying the same criteria to fainter stars does appear to significantly increase contamination), but it does appear to have greater sensitivity to the populations with ages of up to a few tens of megayears.

Most of the selection criteria devised in this work preferentially target low-mass YSOs, still in the pre-main-sequence (PMS) phase of stellar evolution. Young late B, A, and F stars reach the main sequence quickly and become difficult to separate from field stars using conventional photometric selection criteria. To identify them, it is however possible to take advantage of the fact that young stars generally form in large associations, typically with hundreds or thousands of other members. Young populations tend to be dynamically cold, with velocity dispersion of less than a few kilometers per second. Thus, it is possible to find young moving groups by performing a clustering analysis in position and velocity phase space. As a selection of likely members using this method does not have a dependence on the spectral type, it makes possible to include young B, A, and F stars alongside later type stars.

Kounkel et al. (2020) have applied hierarchical clustering on Gaia DR2 data within 3 kpc of the Sun. The initial data selection consisted of π > 0.2 mas, −30 < b < 30°, ${v}_{\alpha ,\delta }^{\mathrm{lsr}}\lt 60$ km s⁻¹, as well as additional cuts based on astrometric and photometric quality. The clustering was performed with HDBSCAN (Campello et al. 2013) in several slices in distance and then stitched together. In total more than 8000 moving groups were identified consisting of ∼1 million stars. The ages of the identified moving groups were estimated through an isochrone fitting using a neural net Auriga (Kounkel et al.2020).

We selected all sources in the moving groups with an age $\mathrm{log}t(\mathrm{Myr})\lt 7.5$. The resulting subset forms the basis of YSO_Cluster carton, introduced in V0 version of targeting, and split into YSO_Cluster_APOGEE and YSO_Cluster_BOSS in V0.5, containing 45,461 and 59,065 stars respectively.

Older populations, with $\mathrm{log}t(\mathrm{Myr})\gt 7.5$, are being considered by the survey as a part of an open fiber program, but only as targets of opportunity, with a single epoch obtained with either BOSS or APOGEE, and are not included among the core programs of the survey.

There have been several dedicated studies that focused on identifying PMS stars using an H-R diagram generated through Gaia photometry and astrometry. We incorporate the resulting catalogs from two such works.

Zari et al. (2018) have selected low-mass PMS stars using Gaia DR2 data within 500 pc that are found above (and therefore are younger than) the 20 Myr PARSEC isochrone (Marigo et al. 2017) and fainter than M_G > 4 mag. The photometry has been first extinction corrected, excluding sources with A_G < 0.92 mag, to avoid reddened field stars. Furthermore, additional cuts have been made to select stars with low parallax errors (σ_π /π < 20%), to limit the sample to disk stars (total tangential velocity $\sqrt{{v}_{l}^{2}+{v}_{b}^{2}}\lt 40$ km s⁻¹), and to produce a "clean" H-R diagram as suggested by (Lindegren et al. 2018, Appendix C).

This selection is effective for nearby populations, but at larger distances, the sample becomes strongly contaminated by field stars, both due to an imperfect match of real photometry to the isochrones, and due to imperfections in the extinction correction. An alternative approach was considered by McBride et al. (2021), using a neural network Sagitta. It was trained on Gaia and 2MASS photometry of stars in young populations from Kounkel et al. (2020) to autonomously identify low-mass PMS stars, automatically adjusting the color–magnitude threshold based on age and distance of a star. The constructed sample extended up to π > 0.2 mas. It consisted of the sources with a classification probability >70%, and it also had several data quality criteria, such as σ_π /π < 10% or σ_π < 0.1 mas, precision in Gaia photometry in all bands <10%, recommended cuts based on the photometric excess noise, as well as Gaia RUWE<1.4.

These two catalogs form the basis of YSO_PMS_APOGEE and YSO_PMS_BOSS cartons that were first introduced in V0.5 version of targeting, containing 76,332 and 73,213 stars respectively. Initially, both catalogs used Gaia DR2 data; in V1 the catalog from McBride et al. (2021) was upgraded to EDR3 version, which (due to magnitude limits) did not change substantially, except for minor improvements in sensitivity to more distant PMS stars.

Most of the YSO cartons (with exception of Cluster and CMZ) focus exclusively on low-mass YSOs, as they are most distinct from field stars. Intermediate and massive stars, on the other hand, reach the main sequence very quickly, and thus become difficult to differentiate.

However, as OB stars have short lifetimes, they would always be young. Thus, to fill the gap in targeting, we selected sources based on examining the placement of known OB stars from Maíz Apellániz et al. (2016):

• 1.  
  −0.2 < (G_BP − G_RP) < 1.1;
• 2.  
  M_G < 1.6(G_BP − G_RP) − 2.2;
• 3.  
  G < 18 mag, π > 0.3 mas.

In V0, this selection formed the basis of YSO_OB carton, which was split into YSO_OB_APOGEE and YSO_OB_BOSS, both containing 8670 stars. However, as all of the selected stars are very bright (typically G < 12 mag), they currently cannot be observed with BOSS without offsetting their positions, due to the saturation limit of the instrument.

In V0.5, these cartons were rendered obsolete, as all of the targets that are a part of YSO_OB are a perfect subset of OBA_CORE program within SDSS-V (Zari et al. 2021); thus, they do not require duplication of efforts from multiple programs.