THE NORTH AMERICAN AND PELICAN NEBULAE. II. MIPS OBSERVATIONS AND ANALYSIS

The MIPS observations of the NAN complex were conducted in two epochs. The first (and largest) component took place on 2006 June 10–11 and covered ∼6 deg² in the region of highest extinction (see additional discussion in G09). These observations were part of Spitzer program id 20015 (PI: L. Rebull); the seven AORKEYs⁷ appear in Table 1. Additional observations were obtained on 2008 June 24 as part of program 462 (PI: L. Rebull) in an effort to increase the legacy value of this data set. They were designed to cover the corner of the map to the same integration time as the main map and added about a square degree to the total coverage. These two additional AORKEYs also appear in Table 1 and are referred to as the "patch data."

The center of the final ∼7 deg² map is (J2000) right ascension (R.A.; α), declination (decl.; δ) = 20:54:18, +44:05:00 (313575, 440833), or galactic coordinates l, b = 848, −053, or ecliptic coordinates (J2000) 3379, +577. Since this region is at a high ecliptic latitude, the field of view rotates by about a degree per day. In order to make AOR coverage as independent as possible without constraints, map offsets (see Table 1) were used along with a group-within constraint to ensure even coverage and minimal overlap between tiles. Also as a result of the high ecliptic latitude, asteroids are not a significant concern; we have just one epoch of observations per field. Figure 1 shows the region of three-band coverage with MIPS, and the four-band coverage with IRAC (discussed in G09). There are about 6 deg² with both MIPS and IRAC coverages.

Zoom In Zoom Out Reset image size

Medium-rate MIPS scan maps were obtained with the spacing between adjacent scan legs set at 160'', 50% of the 24 μm detector width. The total integration times were 84 s, 42 s, and ∼4 s at each point in the map, respectively, for 24, 70, and 160 μm. These observations therefore can detect photospheres at 24 μm down to about 3 M_☉ at the distance of the NAN complex. Much lower-mass protostars with infrared excesses due to a circumstellar disk can also be detected; typically, we can detect down to ∼0.3 M_☉ (mid-M) stars with IR excesses. IRAS data, though incomplete in this region (see Appendix A), show 100 μm emission peaks up to 1800 MJy sr⁻¹, so we did not expect to retrieve valid (unsaturated) 160 μm data, and thus did not plan on complete coverage in that band. However, these data (both the original and the secondary "patch" data) were taken in cold MIPS campaigns and, although undersampled and saturated in places, the smoothed 160 μm data are still valid representations of the emission, see below.

The IRAC observations and data reduction are discussed in G09. We note that since the publication of that paper, we identified a minor error in our analysis where some objects were erroneously identified as saturated at 8 μm, and the total size of the map was overestimated. In the discussion below, when the G09 selection methodology is included, it is meant to be the G09 selection methodology as imposed on the corrected catalog.

We note for completeness that the four channels of IRAC are 3.6, 4.5, 5.8, and 8 μm, and that the three channels of MIPS are 24, 70, and 160 μm. These bands can be referred to equivalently by their channel number or wavelength; the bracket notation, e.g., [24], denotes the measurement in magnitudes rather than flux density units. Further discussion of the bandpasses can be found in, e.g., the Instrument Handbooks, available from the Spitzer Science Center (SSC) Web site.

Figures 2–4 show the individual mosaics by channel, and Figure 5 shows a three-color image with all three MIPS channels included. To guide the eye, the labels for IC 5070, L935, and NGC 7000 appear in the same location in Figures 1 and 2 (optical and 24 μm); the bright emission contours on the upper left and lower right seen in all three MIPS bands are indicated in Figure 1. There is substantial extended emission in all three MIPS channels throughout the MIPS maps; see Section 2.3.

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

2.2.1. 24 μm Mosaics and Point-source Photometry

2.2.2. 70 μm Mosaics and Point-source Photometry

2.2.3. 160 μm Mosaics

2.3.1. Morphology

Bright, complex, extended emission is present at all three wavelengths throughout the MIPS NAN mosaics (Figures 2–5).

At 24 μm, the diversity of textures found in the ISM is stunning. There are dark lanes in the Gulf of Mexico and north of the brightest part of the Pelican (see Section 6); there are broad arcs near 31425, 4493 (20:57:00 +44:55:48) that somewhat resemble the ISM in the Pleiades; there are striations near the center of the map that seem to indicate large arcs centered on the Gulf of Mexico; there are highly localized nebulous very bright blobs; and, finally, there is overall turbulent ISM structure that appears to be illuminated by a source near the center of the map. The brightest portions so illuminated are to the northwest (forming the neck of the Pelican), and the southeast (forming the west coast of Mexico, the "Mexican Riviera"); these edges are indicated in Figure 1 for comparison.

Much of the brightest structure seen at 24 μm is also seen at 70 μm. Several of the highly localized nebulous very bright blobs are still resolved at 70 μm. The brightest portions of the ISM at 70 μm are not only the Pelican's neck and the Mexican Riviera, but also the region near several "Carribbean islands." The 24 μm striations seen in the middle of the map and to the northeast are not apparent in the 70 μm map.

At 160 μm, the brightest parts of this map are the Pelican's neck and two of the highly localized nebulous very bright blobs. The Mexican Riviera and the Carribbean islands are still discernible, even in the highly smoothed map. As in other star-forming regions, the 160 μm emission closely follows the A_V extinction contours, as derived by Cambrésy et al. (2002). Plotting A_V values against 160 μm surface brightness, Figure 6, yields a result similar to other star-forming regions. As in molecular clouds such as Taurus (Flagey et al. 2009), the one-to-one correlation holds to A_V∼5 mag, perhaps higher. For large A_V, we see the "saturation effect" of the near-IR measurements, i.e., the near-IR does not accurately portray the dusty/cold ISM beyond A_V∼10.

Zoom In Zoom Out Reset image size

In all three MIPS maps, the darkest regions are to the southwest and far east, where little ISM and few point sources at 24 μm (none at the other MIPS bands) are apparent. To the southwest, since the star counts at IRAC bands are not dramatically higher in this region, we suspect that we are not seeing the edge of the cloud, but perhaps a part of the cloud that has not yet begun star formation in earnest. The farthest east, darkest region seen in MIPS does not have IRAC coverage, though we suspect that the map extends off the edge of the cloud here.

2.3.2. Comparison to NGC 7023

We can compare the colors of the entire NAN region with those of the classical reflection nebula/photo-dissociation region (PDR) NGC 7023 (e.g., Werner et al. 2004a).

We assume ∼5% uncertainties in the measurements themselves, dominated by the absolute calibration. The overall 8:24 μm ratio is quite similar for the NAN (0.414) and NGC 7023 (0.444), as is the 24:70 μm ratio (NAN, 0.280; NGC 7023, 0.234). We interpret this as that the relative number of polycyclic aromatic hydrocarbons (PAHs) to "Very Small Grains" (VSG; ∼1.2–15 nm)⁹ is similar in the two regions and that the NAN's properties are similar to that of a PDR (consistent with the image morphology).

The 70:160 μm ratio, however, is much different—0.320 for the NAN and 0.546 for NGC 7023. Even if the uncertainties are ∼20%, these values are still significantly different. This suggests that colder dust (inferred from the stronger 160 μm continuum) arising from "Big Grains" (BG) in the NAN brings down the ratio. This seems to be consistent with the "voids" of emission in the images of the region, which could indicate even colder dust, particularly on the southwest corner of the image. This kind of behavior can be seen in Herschel images (e.g., Paradis et al. 2010).

Figure 7 compares the IRAC 8 μm data with the MIPS 24 and 70 μm data. In this composite, similar structures are seen at 8 and 24 μm, as in a PDR. One of the most remarkable things of this composite image is the "veil" or wispy structure at 24 μm in the middle of the image without a counterpart from the dust tracers. Perhaps this is due to fine structure emission lines (e.g., [Fe ii] at 26 μm). This idea is reinforced by the small bowshock at R.A. ∼ 31425 and ${\rm{decl.}}\sim 43\mbox{$.\!\!^{\circ} $}1$ (20:57, +43:06).

Zoom In Zoom Out Reset image size

2.3.3. FCRAO Data

Co-authors J.M.C. and L.A.H. obtained Five College Radio Astronomical Observatory (FCRAO) data of this region in 1998. A complete discussion of these data is beyond the scope of the present paper; Figure 8 shows that the 160 μm emission and the ¹³CO emission are well correlated.

Zoom In Zoom Out Reset image size

The IRAC map (G09) covers an area comparable to that covered by MIPS (see Figure 1). G09 discusses the IRAC source extractions and bandmerging with data from the 2MASS (at 1.25, 1.65, and 2.17 μm; Skrutskie et al. 2006). In summary, the IRAC and 2MASS point sources were matched by position with a matching radius of 2''. The zero points we used for conversion between magnitudes and flux densities are 1594 Jy, 1024 Jy, and 666.7 Jy for JHK_s, respectively; for IRAC's four channels in order, they are, 280.9, 179.7, 115.0, and 64.13 Jy.

G09 also describes matching to our BVI_c optical catalog, which is based on CCD imaging for the ∼2° × 2° region in the heart of the NAN complex. There is extracted photometry for about 19,000 stars. This BVI_c photometry is included in the spectral energy distributions (SEDs) and discussion here. The zero points we used were 4000.87, 3597.28, and 2432.84 Jy for B, V, and I, respectively.

To match the MIPS data to the IRAC data, we have first bandmerged the 70 μm detections to the 24 μm detections using a 10'' merging tolerance. By individual inspection, all the 70 μm sources have 24 μm counterparts. We assigned the MOPEX-derived 24 μm coordinates to each of the sources. Then, we merged the 24 + 70 μm catalog to the IRAC+2MASS catalog using a tolerance of 2'', a value determined via histograms of nearest multi-band matches and experience with other star-forming regions (see, e.g., Rebull et al. 2010). (We note that the absolute pointing accuracy within a single 24 μm observation is ∼14, and the internal positional accuracy within a 24 μm observation can be much smaller than that, but we have empirically found that, once multiple observations are combined and compared to other bands, 2'' is the best value to use.)

Since G09, we have also included data from the Initial Data Release from the INT Photometric Hα Survey of the Northern Galactic Plane (IPHAS). The IPHAS survey was carried out in narrowband Hα and Sloan r' and i' filters, down to red magnitudes fainter than 20 (see Drew et al. 2005 for a general survey description). For the purposes of adding these points to our SEDs, we took the effective wavelengths for Hα, r', and i' to be 0.656, 0.624, and 0.774 μm, and the zero points to be 2974.4, 3173.3, and 2515.7 Jy, respectively.

There are IRAS data over parts of out map (though it is in one of the missing "wedges"), and both MSX and AKARI data over the entire region. We considered including these data for analysis of YSO candidates, but the source density is high, and these prior missions have low enough resolution and are shallow, so we chose not to include these data in, e.g., SEDs. However, we did compare our results to those from these missions; see Appendix A.

We also included literature data from smaller (e.g., not all-sky or Galactic Plane) surveys; see the next section.

There are several surveys of the NAN in the literature, and we compare our catalog to the literature to (1) include additional optical and near-IR data where possible, (2) identify known YSOs, and (3) identify known non-YSO (contaminant) objects. We then used the population of known YSOs and non-YSOs to inform our color-based selection of new YSOs. We now enumerate the literature surveys we included on a case-by-case basis. Where necessary (usually the very old surveys), we used original published finding charts and 2MASS to obtain updated coordinates. Some objects cannot be updated in this fashion, as there are multiple close-by objects of comparable brightness. In a few cases, we believe that certain sources from the literature do not really exist, as detailed below. (Note that prior identifications appear in Tables 3 and 9, discussed below.)

As part of a large Mount Wilson survey for Be stars, Merrill & Burwell (1949, 1950) identified four Be stars in this region based on bright Hα lines as seen on objective-prism photographs. Three of the four stars are in our maps, and all three are very bright at Spitzer bands. Just one (MWC 1032 = HD 198931) is not saturated at all four IRAC bands.

Among the first to specifically and systematically study the NAN complex itself, Herbig (1958) surveyed 6 deg² and identified 68 stars with Hα in emission, which he took to be indicative of youth. Many of these objects subsequently have become famous objects, including well-studied Herbig AeBe (HAeBe) and FU Orionis stars. Herbig's paper contains several prescient observations: (1) that the part of the complex identified as NGC 7000 is just a segment of a much larger reflection and emission nebula, (2) the exciting star(s) is (are) difficult to identify, (3) the Gulf of Mexico region may contain the exciting star of the H ii region, and contains several young and very red stars. With respect to (1), the entire complex with the nebulae is now usually regarded as part of the same molecular cloud (see, e.g., Feldt & Wendker 1993; Feldt 1993). For (2), Herbig (1958) considers and rejects several stars as the exciting source of the nebula. He concludes, agreeing with Osterbrock (1957), that the source of excitation is likely behind the Gulf of Mexico. In regards to (3), five of the stars in the Gulf of Mexico region were bright enough for Herbig (1958) to identify, and he suggested that these objects may be "deep in, or behind, the dark cloud." Additionally, as described above, Herbig (1958) identified a cluster of young objects near IC 5070, leading him to conclude that this region was the most active region of star formation in the complex. We have used 2MASS and the finding chart in Herbig (1958) to obtain updated coordinates for these objects, and they are listed in our Tables 3 and 9 as "IC 5070 cluster N." Other objects identified in that paper usually appear in Tables 3 and 9 as "LkHα" numbers, as per the convention. We note here that LkHα 190 may be extended at Spitzer bands.

The next large, systematic survey of the NAN complex is Welin (1973), who, like Herbig (1958), looked for stars bright in Hα. Welin (1973) surveyed a larger region than Herbig (1958), finding 141 objects down to V∼ 16, just 35 of which were previously known. Spectral types for several of these objects were also reported. These objects appear in Tables 3 and 9 as "UHa" numbers when no more common identification is known.

In the ground-breaking paper by Cohen & Kuhi (1979), among observations for many other clusters, some observations are reported for objects in the NAN complex. They identify two new infrared objects (IRS 1 and IRS 2) in the NAN complex, along with two loose groupings of stars bright in the infrared: a group near LkHα 188 (in the northern Gulf of Mexico region), and IRS 3, 4, 5, and 6 (near LkHα 185). These observations were difficult, obtained with a single-pixel detector drift scanning across the region. By comparison to the Palomar Observatory Sky Survey (POSS), 2MASS, and our Spitzer images, we can recover the stars identified as LkHα 188/G1-5 (G2 and G5 are multiples), though the stars are close enough together that extracting photometry for each object individually using our current pipeline is not always possible due to resolution concerns at 8 μm and longer. Of the remaining objects identified by Cohen & Kuhi (1979), we cannot recover IRS 1-6. While of course it is possible that YSOs can vary substantially in the IR, our observations are significantly more sensitive than these early observations, and a dimming of many orders of magnitude would be possible before we would not recover them. We conclude that these objects do not really exist. The objects that do exist appear in Table 3 (and Table 9) using the same notation as Cohen & Kuhi (1979).

In the process of obtaining CO observations of an expanding molecular shell in this region, Bally & Scoville (1980) reported on 11 bright infrared sources. Using 2MASS and our data, we obtained updated positions for these objects and included them in our database, comparing them to subsequent literature as well. We note that these objects are all very bright in IRAC, often saturated. Objects 1, 7, and 8 have been identified by Comerón & Pasqali (2005; see below) as asymptotic giant branch (AGB) stars, and thus are not YSO candidates. Bally and Scoville's object 2 is in the neck of the Pelican, and resolves with MIPS into bright nebulosity and multiple IR sources; these are likely young objects. Object 5 is a binary (one component of which is identified as an F3 by Straižys et al. 1993). Finally, object 9 appears to be the same as Object 1.

Ogura et al. (2002) reported on 32 stars bright in Hα near bright-rimmed cloud 31, which is near the Pelican Nebula. Some of these objects (7, 8, 10, 15, 18, 19, 20, 22, 28) are in regions that are very bright in the Spitzer bands, and therefore may not be recoverable at some or all Spitzer bands. Some (9, 19) are not recovered with Spitzer. One (25) may be linked to more than one IRAC source.

Villanova et al. (2004) studied NGC 6997, which is not at the distance of the NAN complex, but is covered by our map; see Appendix B.1 below. (Note that this cluster has been commonly confused with NGC 6996; see Laugalys et al. 2006b and Corwin 2004.) Villanova et al. published UBVRI photometry for 2700 stars in this region, which we have matched to objects in our catalog as optical supporting data, in addition to whether or not the objects are members of NGC 6997. The Villanova et al. (2004) optical survey reaches fainter objects than our survey does; ∼1200 objects from this survey do not have counterparts in our survey.

Comerón & Pasquali (2005), in their quest to find the exciting star of the complex, studied 19 objects, six of which they determined to be AGB stars, and five more of which they determined to be carbon stars; the remaining objects were candidates for the exciting star, e.g., young high-mass stars. Both the young high-mass stars and the objects confirmed not to be young high-mass stars are useful to include in our catalog. We note here that their Object 14 appears to resolve into at least three Spitzer sources. Comerón & Pasqali (2005) identify Bally and Scoville's Object 10 as the exciting source for the nebula.

We included in our database the spectral types for several NAN complex objects that were reported as part of large multi-cluster studies; Shevchenko et al. (1993) monitored HAeBes in many clusters, and Terranegra et al. (1994) obtained Strömgren photometry of Orion-population stars. Here, and in general for this database, spectral types of objects obtained via spectroscopy were allowed to supercede spectral type estimates obtained via photometry.

The group at Vilnius Observatory in Lithuania has embarked on a multi-decade campaign to obtain multiband photometry (from which they derive estimates of reddening and spectral types) for thousands of stars in star-forming regions, including hundreds of stars in the NAN complex. The Vilnius bands of observation are U (0.345 μm), P (0.374 μm), X (0.405 μm), Y (0.466 μm), Z (0.516 μm), V (0.544 μm), and S (0.656 μm). Photometrically derived spectral types (and CCD photometry where available) from all of these papers were included in our collection of known objects: Straižys et al. (1989, 1993, 1999, 2008), Zdanavičius & Straižys (1990), Laugalys & Straižys (2002), Laugalys et al. (2006a, 2006b, 2007), Straižys & Laugalys (2008a, 2008b, 2009), and Corbally et al. (2009). Note that, in addition to photometry and spectral types, Laugalys et al. (2006a) and Corbally et al. (2009) specifically identified several YSO candidates as having strong Hα emission. The adopted zero points, kindly provided by V. Straižys (2010, private communication) are, in the order UPXYZVS, 1325, 2112, 4106, 4169, 3949, 3748, and 3141 Jy.

The IPHAS team, in Witham et al. (2008), identified bright Hα sources in the first 80% of their survey, which includes the NAN region. The IPHAS team did not attempt to identify the nature of the sources, simply that they were bright in Hα. Since YSOs are expected to be bright in Hα, the ∼80 stars from that catalog in this region were also matched to our catalog, and the information that they are bright in Hα is folded in to our assessment of YSO candidacy below. We note that some stars with Hα in emission could be active stars or classical Be stars with gaseous disks, not necessarily young stars.

Finally, G09 identified 1657 YSO candidates based on IRAC data. Since then, we have continued to improve our catalog quality (most notably identifying some objects incorrectly marked as saturated at 8 μm), and, using the same algorithm as described in G09, we can now identify 1750 YSO candidates based just on the IRAC NAN data. Objects identified using the G09 methodology but not rediscovered using the MIPS-based search here are listed explicitly in Appendix C.

In summary, we merged our Spitzer-based catalog (from Section 2.4) to these existing data from the literature. There is literature information for about 3600 stars scattered throughout the map, including photometry and spectral types or other classification (such as if they were oxygen- or carbon-rich AGB stars, or red clump giant stars), or even just prior identifications (such as "GSC N" or "TYC N," listed in the tables below). Thus, our final multi-wavelength catalog includes MIPS, IRAC, 2MASS, IPHAS, and other optical data, both from the literature and that which we ourselves have obtained. After including all of this ancillary data, there are spectral types for more than 1200 objects in our catalog, ranging from O5 to M8, with ∼400 additional generic AGB, carbon star, and red clump giant identifications. These types are largely photometric (as opposed to spectroscopically obtained), and largely for types K and earlier.

Just 142 previously identified objects of any sort are missing entirely from our Spitzer-based catalog (aside from those faint ones from Villanova et al. 2004); the overwhelming majority of those are saturated at all Spitzer bands (and thus do not appear in our catalog) or resolve into multiple pieces when viewed with Spitzer (and thus are not formally matched to a single Spitzer source).

Of the previously identified YSOs in the literature (identified in the tables below), it is interesting to consider what fraction are present in our database. About 200 objects in our catalog are identified specifically as previously known YSOs, though this is likely to consist of a mixture of true members and interlopers. Just ∼50% of them have measured flux densities in any MIPS band, and 95% of them are measured at any IRAC band. Most of the "missing" ones are saturated or in very crowded regions in the Spitzer images, which is consistent with the previous surveys being shallower and/or lower spatial resolution. (Note that Table 5 in the discussion below includes statistics on previously known objects recovered or not using G09 and the primarily MIPS-based method presented here.)

Table 2 presents MIPS-focused statistics on the ensemble catalog spanning 0.35 μm through 70 μm. About 75% of all the MIPS-24 sources have an IRAC counterpart. However, the IRAC and MIPS maps do not cover exactly the same area. Considering just the region of the MIPS map where there are also IRAC data from all four bands, ∼85% of the 24 μm sources have a measured counterpart at at least one IRAC band (where the 15% without an IRAC counterpart are overwhelmingly fainter than ∼8th mag, and thus clearly on the very faint tail of the distribution; see Figure 9 below); 60% of the 24 μm sources have a measured IRAC flux density at all four IRAC bands. There are near-IR JHK_s data covering this entire region from 2MASS, but 2MASS is relatively shallow; 69% of the MIPS-24 sources have a 2MASS counterpart. (We note that this region has also been covered by the UKIRT Infrared Deep Sky Survey (UKIDSS) to much deeper limits than 2MASS; we will include these data in a forthcoming paper.) The optical data are highly affected by reddening; 45% of all the MIPS-24 sources have an optical counterpart of any sort (from our data, IPHAS, or the Vilnius photometry).

MIPS sources in general are a combination of cloud members, foreground/background stars in the Galaxy, and the extragalactic background. Since the NAN complex is essentially in the Galactic plane (b ∼ −0.6) and is viewed along a spiral arm, we expect to see mostly Galactic sources, with few extragalactic objects, and we expect an excess over the Galactic background attributable to cloud members. Figure 9 shows the observed NAN 24 and 70 μm differential source counts in comparison to observed source counts from the 6.1 deg² Spitzer Wide-area Infrared Extragalactic Survey (SWIRE; Lonsdale et al. 2003) ELAIS (European Large Area ISO Survey) N1 extragalactic field¹⁰ (the cores-to-disks (c2d; Evans et al. 2003, 2009) reduction is used here, as in Rebull et al. 2007, and the distribution is scaled for relative areas but not additionally reddened). The sources that we see in the NAN do not resemble the source count distribution for extragalactic sources. For 70 μm, scaled source counts from the ∼8 deg² Vulpecula OB association (Vul OB1; Billot et al. 2010) from MIPSGAL (Carey et al. 2009) are also shown for comparison. Like the NAN, Vul OB1 is in the Galactic plane (l∼60.5, b∼0), the sky background levels are comparable to those in the NAN, and the Spitzer observation covers an appreciable area such that there are enough sources to effect a comparison. The source counts in the NAN much more closely resemble that from a star-forming region like Vul OB1 than they do an extragalactic region like SWIRE. While the shapes of the differential source count distribution are similar between the NAN and Vul OB1, there are more sources overall, and more bright sources, in Vul OB1. In both cases, we have comparable data and data reduction strategies, and we are looking along a spiral arm. In the case of Vul OB1, the star-forming region is 2.3 kpc away, which is considerably further away than the NAN. As such, the Vul OB1 region encompasses more foreground volume, and many more foreground sources are included in the sample. (More evidence for this can be found in the fact that most of the NAN 70 μm sources are in the Gulf of Mexico region, as opposed to more evenly distributed over the field.)

Zoom In Zoom Out Reset image size

In order to begin to place constraints on whether we see a higher count rate of MIPS sources over the generic Galactic background (e.g., cloud members), we can compare the source counts within our own maps. The portion of the map to the northeast can be seen to be very likely to be "off-cloud" by inspection of the cloud morphology at all available bands (optical through 160 μm). The general sky brightness is lower at all bands in this region; see Figures 2–5. There are just five 70 μm point sources in this very roughly 1 deg² region; extrapolating this contamination rate to the rest of the map implies that ∼35 of the NAN 70 μm sources may be generic Galactic background, e.g., contamination. At 24 μm, in the same region, there are ∼400 sources, implying that as many as ∼2500 of the NAN 24 μm sources may be Galactic contamination, or roughly half of all of the detected point sources. However, this is likely to be a strong function both of source brightness and location in the map, and moreover sources that have infrared excesses are more likely to be cloud members than background, as discussed below.

Galactic stars contribute substantially to the MIPS source counts in NAN. This is different than many of the other nearby star-forming regions studied with Spitzer (see, e.g., the series of papers from the c2d Legacy team, such as Rebull et al. 2007 for MIPS observations in Perseus; also see Rebull et al. 2010 for Spitzer observations in Taurus) which are out of the Galactic plane. The contaminants of greatest concern to us are the ones that appear to have YSO-like colors. These include some AGB stars (oxygen- and carbon-rich); although intrinsically very bright, such that they would be saturated even at large distances, once they are reddened by the NAN cloud, they can appear fainter and have thus have colors and overall brightnesses suggestive of YSOs (see, e.g., Robitaille et al. 2008).

Some AGBs (including specifically carbon stars) are already known in the direction of the NAN; see Section 2.5 above and Figure 11 below. Although this is far from a complete sample, we can examine the number counts of these objects in our map in an effort to quantify how many AGB contaminants we might have. Twenty-five of these objects are seen in 24 μm, and two are seen at 70 μm as well. Their colors suggest that they could indeed be confused with bright YSO candidates. All of these objects are relatively bright, e.g., have [3.6] between 6.6 and 10.7 mag, but this is likely a bias in that the previous studies tended to focus on the brighter objects.

We can also attempt to constrain the AGB contamination by comparison to other regions studied with Spitzer. Harvey et al. (2006) argue that any AGB star in our Galaxy is likely to be saturated in most IRAC observations (except for cases of very large A_V). Rebull et al. (2007), using data from Blum et al. (2006) obtained in the Large Magellanic Cloud as an aid to quantifying the AGB contamination in Perseus, find that a typical AGB star (in the direction of Perseus) would appear so bright as to be saturated in the 2MASS survey, but not with MIPS. Harvey et al. (2007a) argue that, in their 0.58 deg² field in Serpens (at b ∼ 5 deg), they expect up to six AGB stars and estimate that they actually find three. Scaling from those values to our ∼6 deg² field, we might expect ∼30–60 AGB contaminants. However, in spectroscopic follow-up of half of the Spitzer-selected YSO candidates from Serpens, Oliveira et al. (2009) found a much higher AGB contaminant rate among the Spitzer-selected YSO candidates—25%. Unlike Perseus, where we are looking out of the Galaxy and so there is an "edge" to the distribution of AGBs, both the NAN and Serpens have sightlines in the Galactic plane, where extinction cuts off the AGBs before the Galactic population really falls off. Extinction makes the AGBs look even more like YSO candidates, because it can make an object selected from a single color–magnitude diagram (CMD) appear to have an IR excess, and more generally because it makes the AGBs look fainter, comparable to YSOs in the cloud. Thus, a high contamination rate among the YSO candidates from AGB stars is expected, probably anywhere from the ∼5% estimated originally by a scaled Harvey et al. (2007a) to the ∼25% based on the Oliveira et al. (2009) spectroscopic follow-up of half the Spitzer-selected Serpens YSO candidates.

While of course any individual YSO candidate we find will require spectroscopic follow-up to confirm its YSO status, we conclude that even in the worst contamination case, the majority of objects selected as having red MIPS colors that we report here as YSO candidates are indeed likely to be young stars associated with the NAN cloud complex. Statistically, other properties (e.g., position in multiple CMDs, clustering) make it less likely that a given object is a contaminant, and more likely to be a YSO, as we discuss in the next section.

Of the MIPS bands, 24 μm data are available for the largest number of sources. In order to identify sources with excesses at 24 μm, we compare 24 μm flux densities to those from a shorter wavelength band.

In recent papers such as Rebull et al. (2007) or Rebull et al. (2010), the K_s versus K_s − [24] CMD was used to find YSO candidates. In cases of nearby star-forming regions, or regions where there are no IRAC data, this can be effective. However, there are some good reasons to use [3.6] versus [3.6]−[24] rather than K_s versus K_s − [24]. There is an intrinsic spread in K_s − [24] photospheric colors that is not present in [3.6]−[24] because late-type stars are not colorless at K_s − [24] (Gautier et al. 2007). The effects of reddening are stronger at K_s than at 3.6 μm. And, if 2MASS is the only source of K_s, even short 3.6 μm integrations can reach fainter sources than 2MASS does. In the specific case of the NAN complex, the area covered by MIPS is well matched to that by IRAC, we have substantial reddening, and 2MASS is shallow with respect to the brightnesses expected for late-type young stars at the distance of the NAN (and with respect to the depth of our IRAC data; see Table 2). For the above reasons, we use [3.6] versus [3.6]−[24] as our primary mechanism for selecting YSO candidates.

Figure 10 shows the [3.6] versus [3.6]−[24] plot for the NAN sample in comparison to the 6.1 deg² SWIRE (Lonsdale et al. 2003) ELAIS N-1 and 0.85 deg² c2d Serpens (Harvey et al. 2007b) samples. Ordinary stellar photospheres (likely foreground or background stars) have [3.6] − [24]∼0, and the galaxies as seen in SWIRE make up the large, elongated source concentration near [3.6]−[24]∼6, [3.6]∼15. Objects not in this region, e.g., the brighter and/or redder objects (those to the right of the dotted line in Figure 10), are less likely to be part of the Galactic or extragalactic backgrounds, and more likely to be YSOs with a 24 μm excess. Both the Serpens and NAN clouds have substantial populations of objects meeting these criteria. The distribution of objects found in the NAN is much more similar to that in Serpens than in ELAIS N-1.

Zoom In Zoom Out Reset image size

Reddening can move higher mass objects from the photospheric locus into the region where lower mass YSOs are found. The reddening is far from uniform, with regions of A_V up to ∼30 mag, according to the Cambrésy et al. (2002) extinction map. The extinction is predominantly generated by the clouds in the NAN. As seen in Figure 10, even A_V∼50 does not move objects far enough from the photospheric locus into the region where most of the objects are found in the "red" region of the diagram (to the right of the dotted line). However, the objects that are more moderately red, those immediately to the right of the dotted line, could appear there because of reddening, as opposed to due to an excess due to a circumstellar disk. We can attempt to remove such simply reddened photospheres by incorporating the ancillary data we have collected, see below.

Given the distance to the NAN, and given the background counts analysis above, there are likely to be many legitimate NAN cloud members overlapping the region populated by the SWIRE galaxies, but it is difficult to distinguish the populations given the data we have. For this work, we limit ourselves to the more obvious candidates, e.g., primarily the objects brighter and/or redder than the dotted line in Figure 10, with the caveat that the selection is further modified, here and below. The first modification to this selection is that we look at the significance of the excess. If the excess is seen only at 24 μm, and the measured flux density is less than 5σ above the expected photosphere as extrapolated from a Rayleigh–Jeans slope tied to 3.6 μm, it is less likely to be a real excess source. Conversely, for the objects closest to the photosphere locus (e.g., [3.6]<10), we accept as candidates those objects where the 24 μm excess is more than 8σ–10σ above the extrapolated photosphere, depending on whether the excess is seen at any other bands (e.g., if there is a small excess at 24 μm and a clear excess at IRAC bands, this is more believably an object with a true IR excess than if there is no excess at IRAC bands, just a small excess at 24 μm, and no detection at 70 μm). This has the effect of omitting some sources that technically meet the color criteria and including some sources that are just blueward of the dotted line in Figure 10. There are ∼1200 sources identified from 3.6 and 24 μm data in this fashion.

Figure 11 compares the distribution of various subsets of the NAN data in the [3.6] versus [3.6]−[24] parameter space. Based on our collection of literature data above (Section 2.5), we have a pre-assembled set of NAN members and non-members, which are shown in Figure 11. By and large, these objects fall where we expect them to based on Figure 10—most of the NAN members appear as red in this diagram, which we interpret to mean that they have infrared excesses, and most of the non-members do not. However, it is notable that there are outliers in both cases—some non-members appear to have excesses (largely the AGB contaminants noted above), and there are some members without excesses (which were most likely detected using other mechanisms besides IR colors, such as Hα, although we note that intrinsic stellar variability could also be a contributing factor). Figure 11 also shows the distribution of the YSO candidates selected in G09 as well as the distribution of YSO candidates from our final selection below (Section 4.4). Note that the candidates selected below often recover sources from G09; see Table 5 below and associated discussion. In both cases, the distribution of YSO candidate sources is similar, with the highest density of sources found near [3.6]−[24]∼5, [3.6]∼12.

Zoom In Zoom Out Reset image size

Figure 12 shows [24] versus [24]−[70] for the same three samples as Figure 10, namely, SWIRE ELAIS N-1, c2d Serpens, and the NAN sample. The distribution of sources found in the NAN again is a much closer match to that from Serpens than from SWIRE. Few of the objects seen at 70 μm in the NAN are likely galaxies; several are very red.

Zoom In Zoom Out Reset image size

As discussed above (Section 2.5), we have included in our database a wide variety of ancillary data. These data are not always well matched to our Spitzer data in depth or area. But they can contribute to the assessment of any one object as to its likelihood of being a YSO.

Rebull et al. (2010), working in Taurus, had a similar problem to overcome in that there were ∼44 deg² of Spitzer data, considerable foreground and background contamination, but also considerable ancillary data. The approach taken in that study was to incorporate, in addition to the easily quantifiable Spitzer magnitude and color criteria, qualitative judgments based on what data were available. These included but were not limited to: relative brightness at all available bands, amplitude of excess, shape of SED, apparent (projected) proximity to other previously identified cluster members or non-members, previous identifications, and star counts.

For the NAN data, though the complete catalog is comparable in size to that from Taurus, we have many more objects with YSO-like colors to consider, and different ancillary data. However, our fundamental approach is similar to that in Taurus. We started with the objects selected in the [3.6] versus [3.6]−[24] CMD, and we folded in several considerations to assess the likelihood of youth and therefore membership in the NAN. In each case, the assessment can be described as an educated statistical guess (e.g., an object selected in the [3.6] versus [3.6]−[24] CMD is statistically more likely than an object not selected in that diagram to be a YSO), but that does not ensure that all YSOs will be selected, nor that all objects that are selected are YSOs. Objects that had no available data for any particular one of these items were unaffected by the lack of such data. (Note that the description here includes forward references to columns in Table 4, which tabulates whether a given YSO candidate met the listed criterion.)

• 1.  
  If the object appears to the right of the dotted line in Figure 10, e.g., if it is in the space occupied by most YSO candidates in the [3.6] versus [3.6]−[24] CMD, then it is a likely YSO (Column 5).
• 2.  
  The fainter the object is in [3.6], the more likely that it is a galaxy even if it is red (Column 6).
• 3.  
  If the object is also selected as bright and red (e.g., a likely YSO) in the K_s versus K_s−[24] CMD, then it is more likely to be a YSO, but if those data (K_s and [24]) are available for the target and it is not in the bright and red regime, then it is less likely to be a YSO (Column 7).
• 4.  
  If an object is detected at 70 μm, it is more likely to be a YSO, because the 70 μm survey is effectively less sensitive than the 24 μm survey (e.g., reaches only the brighter excess objects in the NAN—see Figures 10 and 12) (Column 8).
• 5.  
  If the object was found via the IRAC-based method used in G09, then it is more likely to be a YSO; if it could have been found using that method (e.g., observations exist at all IRAC bands) but was rejected, then it is less likely to be a YSO (Column 9).
• 6.  
  Because the complete map we have extends past the limits of the bright nebulosity and presumably past the edge of the molecular cloud, if the object was closer to the middle of the nebula, e.g., within ∼1° radius of the approximate center (α, δ = 313752, 44231 or 20:55:00.5, +44:13:52), then it is more likely to be a YSO than if it is further away (Column 10).
• 7.  
  Because YSOs are more likely to be clustered, if the candidate was in a cluster (see discussion in Section 6.1 below), it is more likely to be a YSO than if it is not in a cluster (Column 11).
• 8.  
  If an object appeared in the literature as a previously identified YSO using any wavelength, it is substantially more likely to be a legitimate YSO. Similarly, if it was previously identified as a likely contaminant, it is substantially less likely to be a legitimate YSO (Column 12).
• 9.  
  In the I_C versus V−I_C CMD, if the object appears substantially above the zero-age main sequence (ZAMS), close to where most of the previously identified YSOs are found, then it is more likely to be a YSO; if the object appears below this (or below the ZAMS itself) it is less likely to be a YSO, at least one belonging to the NAN, unless dominated by scattered light at optical wavelengths. (column 13)
• 10.  
  Similarly, in the IPHAS r' versus r' − i' CMD, if the object appears substantially above the ZAMS (assuming no reddening), near where most of the previously identified YSOs are found, then it is more likely to be a YSO; if it appears below this (or below the ZAMS itself), it is less likely to be a YSO (Column 14).
• 11.  
  If the object was already identified as a bright Hα source in IPHAS (Witham et al. 2008; Column 15), or appears in the r' − Hα versus r' − i' diagram as substantially above the unreddened main-sequence locus (column 16), it is more likely to be a YSO; if the object is below this, it is less likely to be a YSO.

Each object was individually initially assessed resulting in a distribution of confidence levels, which we arbitrarily bin into three groups—∼700 in grade A, ∼500 in grade B, and ∼200 in grade C. However, there is still more information that can be included.

Even with all of the Spitzer+ancillary data above, there are still sources left in the pool of possible candidates that are unlikely to be true YSOs. The SED for each YSO was examined by hand, along with each object's status for each item from the prior semi-automatic step. Objects that were too bright (likely foreground and/or AGB stars) or were clearly reddened photospheres (likely background giants) or had SEDs resembling known extragalactic objects or had very low significance excesses or very sparse data (e.g., not many measurements beyond 3.6 and 24 μm) were removed from the list (∼300 objects) or downgraded in confidence level (∼100 objects). (See, e.g., Rebull et al. 2010 for discussion of similar extragalactic and very bright sources.)

Similarly, there are sources that need to be added to the list of candidates. The color cut in the [3.6] versus [3.6]−[24] diagram above summarily rejected sources in the region of this diagram occupied by SWIRE galaxies, but it is known that YSOs can appear in this portion of the diagram. As discussed above (Section 3), the extragalactic contamination is likely to be low in the NAN region, and moreover, where there is very high A_V, such as in the Gulf of Mexico cluster (see Section 6.2 below), it is unlikely that sources of such red color are extragalactic contaminants. Similarly, there are sources where there is a 24 μm measurement, but the 3.6 μm measurement is missing due to source confusion or reddening or other similar issues; these objects could not be selected in the [3.6] versus [3.6]−[24] diagram but are unlikely to be background galaxies. There are several (∼60) sources like this in the NAN; we have added these sources as grade B or C confidence YSO candidates and indicated them as such in the tables below.

The complete list of 1286 MIPS-selected YSO candidates appears in Table 3, along with Spitzer measurements and, if available, prior identifications and spectral types. Table 4 repeats the entire list, along with their properties as derived here, including the assigned cluster membership, if applicable, the final YSO quality flag (grades A/B/C), the YSO class (see Section 5.2), and which of the above YSO criteria a given object met. In each case (for each individual criterion), "yes" means the object met that given criterion for YSO selection, "no" means it did not, and "no data" ("  ⋅⋅⋅  ") means that it was unable to be assessed for that criterion. For example, Column 5 (abbreviated "sel 3324") indicates whether each object was selected as a YSO candidate in the [3.6] versus [3.6]−[24] diagram ("yes"), or it was not ("no"), or, it was missing a 3.6 or 24 μm measurement, and thus could not be assessed in the [3.6] versus [3.6]−[24] diagram ("  ⋅⋅⋅  "). The column definitions are listed in Section 4.2 above.

As in Taurus, based on the criteria above (Sections 4.2 and 4.3), we have assigned grades to each source to reflect the confidence in its identification as a YSO. There are in the end 511 grade A, 523 grade B, and 252 grade C (with a subset of 23 of those being marked "C−"). The highest confidence YSO candidates generally have at least ∼5 or 6 of the criteria listed above in their favor (yeses in Table 4), and the lowest confidence have ∼3. The cumulative YSO candidate list is plotted in Figure 13 separated by grade. The faintest ones are overall more likely to be contaminants, the clump of bright grade C candidates is some of the ones that were manually downgraded (in Section 4.3) due to less significant excesses, and the ones in the region occupied by SWIRE galaxies are the ones manually promoted in Section 4.3. For the analysis in the rest of this paper, all of the candidates are considered as one large set.

Zoom In Zoom Out Reset image size

Table 5 summarizes a variety of statistics on the YSO sample selected here, that selected by G09 methodology, and the initial sample of YSOs from the literature. Note that, because our MIPS survey is shallower than our IRAC survey for relatively blue objects, ∼40% of the objects selected using the G09 IRAC methodology do not have MIPS counterparts. The objects selected by G09/IRAC but not recovered by our MIPS-based selection are listed in Appendix C. The objects found by our MIPS-based selection but not by G09/IRAC, in most cases, are those with small or no excesses at IRAC bands but clear excesses at 24 μm (and sometimes longer wavelengths as well). Previously identified YSOs from the literature that were not recovered (via IRAC or MIPS selection) do not have an infrared excess that we could measure, and they appear in Appendix C as well. These objects may or may not be legitimate NAN members; we cannot determine this from our data.

In certain contexts below, the objects that were IRAC-selected but not recovered by MIPS, and those previously identified YSOs not recovered by either IRAC or MIPS, may be included in the discussion with the MIPS-selected objects. These instances are all clearly noted.

We have VI_C data for 14% and JHK_s for ∼80% of our YSO candidates. The VI_C objects tend to be the brighter and/or less embedded of the YSO candidate sample.

Figure 14 shows the J−H versus H−K_s color–color diagram for the candidates. The subset of objects with VI_C data are indicated. As suggested by earlier figures, there is generally high A_V in the direction of the NAN. A_V∼10–20 are common. Near-infrared evidence for the disks selected by the Spitzer color criteria is apparent.

Zoom In Zoom Out Reset image size

In the spirit of Wilking et al. (2001), we define the near- to mid-IR slope of the SED, α = dlog λF_λ/dlog λ, where α>0.3 for a Class I, 0.3 to −0.3 for a flat-spectrum source, −0.3 to −1.6 for a Class II, and <−1.6 for a Class III. For each of the YSO candidate objects in our sample, we performed a simple ordinary least-squares linear fit to all available photometry (just detections, not including upper or lower limits) between 2 and 24 μm, inclusive. Note that errors on the infrared points are so small as to not affect the fitted SED slope. The precise definition of α can vary, resulting in different classifications for certain objects. Classification via this method is provided specifically to enable comparison within this paper via internally consistent means. Note that the formal classification puts no lower limit on the colors of Class III objects (thereby including those with SEDs resembling bare stellar photospheres, and allowing for other criteria to define youth). By searching for IR excesses, we are incomplete in our sample of Class III objects.

Figure 15 shows a histogram of the SED slopes, calculated as described, for the MIPS-selected sample, the IRAC-selected sample not recovered by the MIPS-based search, and the complete Spitzer-selected sample (IRAC-selected plus MIPS-selected). The objects not recovered by MIPS in general have more negative slopes, because the objects that are not detected by MIPS are the ones that are the faintest in the MIPS bands, e.g., the ones whose SEDs are falling through the IRAC bands. The IRAC-selected sources with steeply rising SEDs (suggesting that they should have been detected at 24 μm) that are not recovered by our MIPS selection are either off the edge of the MIPS map or saturated at MIPS bands.

Zoom In Zoom Out Reset image size

We repeated the spectral index analysis with slopes fit to just available points between 2 and 8 μm (e.g., ignoring any points at 24 μm and longer). The peaks of the histograms of the MIPS- and IRAC-selected samples move closer together, but the MIPS-selected peak is still to the right of the IRAC-selected peak, and the relative fractions of sources in the Class bins do not change.

The MIPS-selected sample is nearly half Class IIs (48%). Including the IRAC-selected sample as well, the fraction of Class IIs among the selected IR-excess objects rises to 56%. For either sample, there are about 8% Class IIIs, and the remaining objects are evenly split between Class Is and flat objects (with an additional 2% where there are too few points to calculate a slope). The relative numbers are quite comparable to the fraction of such objects found in Serpens by Harvey et al. (2007a) or Perseus (Rebull et al. 2007), suggesting that these regions are all comparable in the status of the circumstellar material and hence possibly their age.

Figure 16 shows the projected locations of all of the YSO candidate sources selected using our MIPS-based selection. Several clusters of objects can be seen by eye here; we will consider clustering formally in the next section. There is also a distributed population. The fact that there is a substantial distributed population here is consistent with results from many other Spitzer studies of star-forming regions, where substantial numbers of YSO candidates are found far from the canonical centers of star formation.

Zoom In Zoom Out Reset image size

In Figure 2, we saw large spatial variation in the 24 μm surface brightness due to the ISM, which renders the limiting flux density a function of position in the image. Similar structures can be seen at the IRAC bands (see G09). It is difficult to assess if this has affected the apparent distribution of the selected YSO candidates. The distribution of YSOs is not uniform, and it is expected based on physical arguments that the distribution of YSOs will be coupled to the ISM, especially at these young ages. We examined the distribution of YSO candidates (and all 24 μm sources) by plotting them on top of the 24 μm image. We do not see a very strong effect of the ISM brightness on our YSO candidate distribution. Qualitatively, we note that the bright "streaks" in the center of the map have few YSO candidates, and also fewer 24 μm sources in general; the even brighter ISM in the region of the Pelican cluster hosts many YSO candidates. We conclude that the ISM surface brightness is not a dominant effect in the structure seen in the YSO candidate distribution. However, many of the YSO candidates are projected against the regions of highest extinction (which correlates well with the MIPS surface brightness), with concentrations near the Gulf of Mexico and the greater Pelican region.

As discussed above, there are a significant number of IRAC-selected YSO candidates that are not recovered by our MIPS-based selection, and these appear in the second panel of Figure 16. Clustering similar to the left panel can be seen by eye, even in this residual IRAC-only selection.

G09 discussed in some detail a formal definition of clustering, using a Kernel method to estimate the density of YSO candidates (note that not the density of sources in the entire catalog) as a function of α, δ. Using the same formalism, we computed a YSO density map of our final set of MIPS-selected YSO candidates. There are three main clusters that can be identified out of this sample. These three clusters appear in the rightmost panel of Figure 16. (A background cluster is also indicated in that figure for reference; see discussion in Appendix B.1.)

We name these three clusters based on their position with respect to the nebular features—the Gulf of Mexico cluster is located in the Gulf of Mexico region of the North American Nebula; the Pelican cluster overlaps with the Pelican's neck, head, and beak; and the third cluster appears above the Pelican's head, hence "Pelican's Hat." We discuss the properties of these clusters in turn below. Cluster membership is indicated for individual objects in Table 4. Ensemble properties are summarized in Table 6; plots of [3.6] versus [3.6]−[24] are in Figure 17 and plots of [3.6]−[4.5] versus [5.8]−[8] are in Figure 18.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Because clustering is considered in the selection of YSO candidates above (Section 4.2), and because some sources were added by hand in regions of high YSO density (see Section 4.3), identifying and defining clusters are somewhat of a recursive problem, with the potential for biased results. We also ran the cluster identification algorithm on the initial and least biased sample of [3.6] versus [3.6]−[24] selected objects (Section 4.1). The three clusters were unambiguously recovered even in this sample, so we conclude that they are independent of the details of our YSO selection algorithm. (We note that heavily reddened K and M giants located behind a very dense dust cloud could appear to be clustered YSOs; see, e.g., Straižys & Laugalys 2008a.)

In the left and middle panels of Figure 16, there are additional smaller clumpings of sources to which one's eye is drawn, most notably a clump on the far west of the distribution. It has a linear string of YSOs in the MIPS-selected sample and a slightly offset clump of sources on the edge of the IRAC map from the IRAC-selected YSOs. Neither of these groupings appear in the clustering analysis above.

There are legitimate YSOs here outside of the clusters, notably the outbursting Class I object PTF10nvg (Covey et al. 2011), which is south of the Pelican cluster.

Parts of two of the clusters described here (the Gulf of Mexico and the Pelican) have been previously identified in the literature (based primarily but not exclusively on Hα); our data suggest expanded spatial definitions (and far more members) for these clusters. The Pelican's Hat cluster is an entirely new discovery based on the MIPS data. In Appendix B, we discuss additional clusters from the literature that are sufficiently unremarkable in the MIPS maps that they are mentioned here solely for completeness.

The SEDs presented below are all in log λF_λ in cgs units (erg s⁻¹ cm⁻²), against log λ in μm.

The "Gulf of Mexico" region is the area of highest extinction in the NAN complex; the Cambrésy et al. (2002) extinction map peaks at A_V∼30, and is likely a lower estimate to the true extinction due to the fact that it was calculated using near-IR data and is lower spatial resolution than our MIPS data. Herbig (1958) identified several stars bright in Hα at the northernmost part of the cluster we describe below. Cambrésy et al. (2002) find three subclusters within this cluster using 2MASS star counts (in their notation, 1, 2, and 3a/b); G09 also finds three subclusters within this cluster using IRAC.

As can be seen in Figure 19, the image of this region is striking at 24 μm (and 8 μm; not shown), with infrared emission and a filamentary dark cloud. Parts of the cloud are dark at 24 μm and even at 70 μm; this region has the highest surface density of 70 μm point sources in the complex. Counterparts at 160 μm can be seen as well, though no viable photometry can be extracted from our 160 μm data. Many jets are visible at the IRAC bands, especially at the northern end of the cluster.

Zoom In Zoom Out Reset image size

The cluster membership is defined as discussed above, and is represented as a contour in Figure 19; that contour is very close to the A_V = 10 contour from the Cambrésy et al. (2002) extinction map. Using the G09 definition of candidate (IRAC) YSOs and clustering metrics, this contour is similar to that for 1500 (IRAC-selected) YSO candidates per square degree. It is about 30' across at its widest part, corresponding to about 4.5 pc at a distance of 520 pc. There are 283 objects seen at 24 μm that we define as being part of the cluster (see Table 6), which is ∼2500 deg⁻². Note that some objects not explicitly selected via the color mechanisms in Section 4 above (such as those objects seen only at 24 μm) have been added to the YSO candidate list on the basis of their projected location in this cluster (and 24 μm flux density). At 70 μm, 10% of the 24 μm sample have counterparts. Most of the objects, 85%, are seen at all four IRAC bands, but 5% are so embedded that they lack any IRAC counterparts at all.

Figures 17 and 18 show the [3.6] versus [3.6]−[24] and the IRAC color–color diagram for this cluster in context with the other clusters discussed here. The Gulf cluster has by far the most 70 μm detections (see Table 6), and the widest range of colors of YSO candidates in both these diagrams. While several objects in the Gulf have extreme IR colors, a major contributing factor to the dispersion in colors is the high A_V.

Our primary YSO candidate selection mechanism above requires MIPS, and we defined clusters based on the location of those MIPS-selected YSO candidates. But, for relatively blue sources, the IRAC survey is more sensitive than the MIPS survey, and there are most likely legitimate Gulf cluster members not detected at MIPS bands. There are ∼4100 IRAC-1 and IRAC-2 sources without MIPS counterparts within the A_V = 10 contour (∼40,000 deg⁻² compared to an average of ∼60,000 deg⁻² over the whole map). Just ∼800 of the objects in this region are detected at all four IRAC bands (∼8000 deg⁻²). Of those objects, just 92 IRAC-selected YSO candidates (requiring selection in all four IRAC bands) are unrecovered by our MIPS-based selection and could be legitimate cluster members. There are additional likely YSOs that we list as YSO candidates based on 5.8 or 8 through 70 μm properties, but they do not have detections at all four IRAC bands. Such sources without 3.6 or 4.5 μm detections would not be selected via the G09 mechanism. The 92 IRAC-selected YSO candidates unrecovered by our MIPS-based selection is the minimum number of additional cluster members. Adding those 92 to the 283 we define (based on MIPS) as the cluster membership, we derive a YSO census of 375 members and a surface density of ∼3500 deg⁻².

To constrain the maximum number of true cluster members, we can look at overall source counts. The number per unit area of four-band IRAC detections (without MIPS detections) on the edges of the map, in regions of much lower A_V, can be taken as roughly the number of background+foreground counts; these are ∼11,000 deg⁻². Considering those objects with IRAC-1 and 2 detections (but no MIPS-24), there are ∼70,000 sources deg⁻² on the map edges. It is thus unlikely that all projected IRAC-only sources could be cluster members, but the IRAC-selected YSO candidates are far more likely to be members than any given IRAC-only source. Additional observations and analysis are needed to distinguish the true cluster members from the foreground stars.

Of 375 cluster members, just 11 were identified in the literature prior to our Spitzer observations as cluster members (with 303 identified in G09). All of the objects known prior to Spitzer are in the northernmost ∼6' of the cluster, and eight of them are in the northermost ∼15 of the cluster, so the size of the cluster is redefined as about 10–100 times larger (in projected area) than previously realized.

Figure 20 shows the projected spatial location of the MIPS-selected objects, coded by their SED slope. The Class I and flat objects are more clustered than the Class II and Class III objects. Table 6 lists the relative fractions of Class I, flat, II, and III MIPS-selected YSO candidates in the cluster. If the candidates found using IRAC only are included, the relative fractions are 32%, 30%, 34%, and 2%, respectively (with 3% for which no slope can be calculated); the largest increase is in the Class II category, consistent with Figure 15. The dense embedded cluster found in Serpens South (Gutermuth et al. 2008) bears some morphological similarity to this region. The mechanism used by those authors to select YSOs is very different than what we use, but they find about 60% Class I+flat objects. In the Gulf, we find that about 62% of all the YSO candidates are Class I or flat (71% of just the MIPS-selected sample). Within the NAN complex, the Gulf has the largest fraction of Class I objects. The overwhelming majority of the NAN objects seen in the [24] versus [24]−[70] diagram (Figure 12) are located in the Gulf. We identify this region in the NAN complex to have the most embedded objects (presumably the youngest?), the highest surface density of YSO candidates, and to be where star formation is actively occurring.

Zoom In Zoom Out Reset image size

Figure 21 shows a sample of 16 SEDs selected out of our proposed cluster membership to represent the range of colors and SED types found in this cluster. One object in this figure is essentially photospheric at all available bands (which may mean that it is a foreground object superimposed on the dense cluster region and erroneously included as a cluster member solely based on its location on the sky), while others in this figure are deeply embedded, appearing only at IRAC+MIPS, not 2MASS or optical. Notably, the object (PTF10qpf) recently identified by Miller et al. (2011) as an FU Ori-like outburst was present in our pre-outburst observations of this cluster.

Zoom In Zoom Out Reset image size

Herbig (1958) identified a cluster of 12 young objects bright in Hα near IC 5070 (the Pelican Nebula). This cluster was denser than the other clusters Herbig identified in the NAN complex, leading him to conclude that this region hosted the most active region of star formation in the complex. While all of these objects are seen at 2MASS and IRAC bands, none are detected at 24 μm, and none have [3.6]−[24] excesses; moreover, none are even identified by G09 as IRAC-selected YSOs. There is, however, still a noticeable clumping of Spitzer-identified YSO candidates in this general region; see Figure 16.

Figure 22 shows the Pelican Cluster region in 24, 70, and 160 μm as well as a POSS-red image. There are many point sources at 24 μm; some sources can be seen in the 70 μm image, but only three can be identified using our pipeline—most are slightly extended or in regions of extremely bright background. No point sources are discernible at 160 μm.

Zoom In Zoom Out Reset image size

The contour defining the cluster membership is shown in Figure 22 as well. Interestingly, unlike the Gulf cluster in Section 6.2 above or in the Pelican's Hat cluster in Section 6.4 below, the definition of the cluster as returned by the formal clustering metrics does not trace an A_V contour. Rather, it links two A_V∼20 peaks, seen in the 24 μm image as the brightest surface brightness regions. It also overlaps with more than half the projected area of two clusters from Cambrésy et al. (2002), clusters 5 and 6, though the regions specifically called out by Cambrésy et al. often include regions too bright and nebulous for many MIPS-24 point sources to have been recovered. Similarly, the G09 clusters 5 and 6 are both located within our MIPS-based cluster contour.

Table 6 lists characteristics of this cluster as we now define it. Forty-eight of the 247 cluster members are previously identified as YSOs in the literature, so we have expanded the likely cluster membership by a factor of ∼5. Herbig (1958) identified 13 more Hα sources in this region, but did not explicitly call them out as being part of the IC 5070 sources he identified as a cluster; Ogura et al. (2002) also found several Hα sources here. In contrast to the 12 original Herbig (1958) IC 5070 sources, we detect these Ogura Hα sources at 24 μm. The fact that nearly 20% of the cluster membership has already been identified in the literature as young stars is consistent with there being overall lower A_V towards this cluster than the Gulf of Mexico and Pelican's Hat clusters identified in this paper. Unlike the Gulf of Mexico cluster, we did not add manually additional objects to the cluster inventory on the basis of 24 μm flux density plus clustering. Because the average A_V is lower here than in the Gulf, the risk of contamination is higher. However, a high fraction of the IRAC-selected (but not MIPS-recovered) sources are likely to be cluster members, as evidenced by the clustering found in G09 (or, for that matter, Cambrésy et al. 2002).

Figures 17 and 18 and Table 6 show that, although it contains objects of a wide range of colors, the Pelican cluster is almost two-thirds Class II objects; in both Figures 17 and 18, there are regions of higher source density. Among the three clusters under consideration here, this one is significantly more tightly clumped in color space or in class distribution, with a much higher fraction of Class IIs and much lower fraction of Class I or flat objects. This is almost certainly related to the relatively lower A_V (and thus less smearing in the color spaces) in the direction of this cluster, but also may be telling us something about the relative degree of embeddedness and possibly the age of this cluster compared to the other two. If this cluster is indeed the oldest of the three clusters discussed here, it has had more time to disperse the surrounding cloud (accounting for the lower A_V) and for the objects to move from their original natal location (accounting for the more spatially dispersed, less tightly clustered locations). Figure 23 shows the (projected) locations of the cluster members. Most of the Class I and flat objects are seen on the east and west edges of the cluster, close to the A_V peaks. Figure 24 shows a sample of SEDs selected out of our proposed cluster membership to represent the range of colors and SED types found in this cluster.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

North of the Pelican's head, in roughly the same location as would be found a hat on its head, is a previously unidentified dark lane at 24 μm, similar to but much smaller than the Gulf of Mexico; see Figure 25. This sinuous lane is about 10' long (∼1.5 pc at a distance of 520 pc), with two bright 24 μm sources at the south end, and several other nearby objects detected at 24 μm. The definition that emerges as part of the cluster formalism above corresponds very closely to an A_V∼8 contour from the Cambrésy et al. (2002) extinction map.

None of the objects found in this region correspond to previously known YSOs or to any other prior identifications. As seen in Figure 25, the region is on the edge of our 70 μm map, and so viable data are obtained only for part of the cluster. Texture in the nebulosity (including some nearby patches not formally included in the cluster definition) can be seen. The 160 μm image has hints of the same structure.

Zoom In Zoom Out Reset image size

Of the three clusters highlighted in this paper, this is the smallest in projected area, number of sources, and surface density. There are 51 sources we define to be part of this region. Table 6 and Figures 17 and 18 suggest that the cluster may be intermediate in YSO candidate properties, between the Gulf and the Pelican clusters. The relative fraction of objects is greatest in the "flat" class, but within Poisson counting statistics, the relative fractions seen in each of the Class I/flat/Class II/Class III categories is indistinguishable from the Gulf of Mexico cluster.

Figure 26 shows the projected spatial distribution of YSO candidates; note that there are no Class III objects (whereas the other two clusters have a few), and that each category of object seems to be smoothly distributed. This may be another indication of youth, or this may mean that the actual cluster is larger than our defined area. Figure 27 shows a sample of SEDs selected out of our proposed cluster membership to represent the range of colors and SED types found in this cluster.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

There are ∼10 IRAC-selected but not MIPS-recovered YSO candidates in this region, and they are almost all located on the edges of our cluster definition. It is difficult to assess if they should be included in the cluster or not, and we have chosen not to include them.

There are at least two objects nearby but not formally in the cluster definition whose morphology in the 24 μm image suggest youth; see Figure 25. About 8' to the northeast of the north end of the dark lane (at 20:51:37, +45:05:15 or 312.9042, +45.0875), there is a resolved patch of 24 μm nebulosity, and ∼55 to the west of the north end of the dark lane (at 20:50:35 +44:57:16, or 312.6458, +44.95444), there is a resolved patch of nebulosity that might be a bipolar nebula, and it is on the edge of a larger circular feature ∼55 in diameter. Since so many of the objects in the NAN are young, we hesitate to add these objects (and their apparently associated point sources) to the cluster definition, but they are likely YSOs, possibly from the same local episode of star formation that prompted the formation of objects in the Pelican's Hat cluster.

IRAS flew in 1983 and observed at 12, 25, 60, and 100 μm. The angular resolution of IRAS ranged from ∼05 at 12 μm to ∼2' at 100 μm; source confusion is clearly a problem in regions like the NAN, where the source density varies from ∼20 arcmin⁻² at IRAC bands to ∼0.2 arcmin⁻² at 24 μm, and the nebular emission is complex at essentially all bands.

Previous results in Perseus from Rebull et al. (2007) comparing MIPS and IRAS data suggest that the IRAS point-source catalog (PSC; Beichman et al. 1988) and faint-source catalog (FSC; Moshir et al. 1992) did reasonably well in identifying true point sources; while the extended emission is complicated in Perseus (as it is in the NAN), the source density is not as high as in the NAN.

IRAS surveyed more than 96% of the sky, but one of those missing sky segments is in the NAN region. There is a missing "wedge" of IRAS imaging data along the eastern edge of our map. Apparently as a result, the point-source extraction routines were not run over much of the NAN map. There are no FSC sources found at all within 4 deg² radius of our map center. There are only PSC sources found in a narrow strip in the western edge of our map, from ecliptic longitudes ∼336° to 337°.

There are 69 PSC objects that overlap our MIPS map, 28 of which are listed as solid detections (quality flag = 3) at 25 μm, and they appear in Table 7. Thirteen of these resolve into nebulosity when viewed at 24 μm; the rest resolve into at least one point source. At 60 μm, there are 28 solid detections (quality flag = 3), just three of which are detected as point sources in our 70 μm map (IRAS 20472+4322, 20467+4340, and 20472+4338); the rest are confused by nebulosity. The nebulosity is complex, and the overall sky brightness is high, in this region.

While MIPS has much more sensitive detectors than IRAS, it also samples a much smaller beamsize, so the achieved surface brightness sensitivity is not tremendously improved over IRAS. Most of the large-scale structure seen in the MIPS maps can also be seen in the IRAS maps. Obviously the tremendous texture found in the MIPS maps is lost in the IRAS maps.

The NAN region is included in the MSX survey of the Galactic plane (Price et al. 2001). A U.S. Air Force satellite, MSX flew in 1996–1997, and observed at 8.28, 12.13, 14.65, and 21.3 μm at a resolution of ∼18''. As for the IRAS maps, hints of the structure found in the ISM can be seen in the lower-spatial-resolution MSX images. The dark lane in the Gulf of Mexico can clearly be seen in the 8.28 μm image (and hints of it can even be seen the 12.13 μm image). The brightest emission regions discussed in Section 2.3 can be seen in the MSX images as well: the Pelican's neck, the Mexican Riviera, and the bright resolved blobs.

About 900 sources are retrieved from the MSX point-source catalog (Egan et al. 1999) for this region. Simply by overlaying them on the Spitzer images, it can be seen that many are clearly confused by the ridges and structure in the nebulosity, as well as point-source confusion itself, though many do correspond to real point sources seen in the 24 μm data. It is somewhat difficult to compare the flux densities quantitatively, as the Spitzer instruments saturate at about the point where the MSX channels are operating at the lower limit of their completeness. There are ∼20 sources in common between MSX Band A (8.28 μm) and IRAC 8 μm, most of which are reasonably well matched in measured flux densities (1σ scatter for the brighter sources is 17%); about 20% of the IRAC observations clearly resolve source confusion in the MSX data. There are ∼40 sources in common between MSX Band E (21.34 μm) and MIPS 24 μm, the brighter half of which are comparable in measured flux densities (1σ scatter for the brighter sources is 37%); about 40% clearly resolve source confusion in the MSX data, most of which are below the 2 Jy completeness limit for the MSX Band E data.

AKARI is a Japanese mission, and it surveyed the whole sky at 9, 18, 65, 90, 140, and 160 μm during its cryogenic phase (2006–2007). With a primary mirror of 68.5 cm, AKARI's resolution is slightly worse than Spitzer's; the spatial resolution is ∼5'' at the shortest bands and ∼90'' at the longest. As of this writing, images from this survey are not available, but catalogs are. We extracted ∼1100 sources in this region from the IRC 9 and 18 μm all-sky point-source catalog (Ishihara et al. 2010) and ∼2100 sources in this region from the FIS 65, 90, 150, and 160 μm all-sky bright source catalog (Yamamura et al. 2009). Not all of those sources are indicated to be high-quality detections at each band; 96% of the IRC catalog is a high-quality detection at 9 μm, whereas just 30% are similar quality detections at 18 μm. For the FIS catalog, 13%, 92%, 31%, and 15% are high-quality detections at 65, 90, 140, and 160 μm, respectively.

By empirical inspection of the Spitzer images, the IRC catalog is not often confused by nebulosity; the overwhelming majority of IRC point sources are point sources for Spitzer as well. Unsurprisingly, the Spitzer data are substantially deeper; the AKARI catalogs only go to ∼0.1 Jy at 9 μm. There are ∼640 objects in common between the catalogs, and the flux densities are reasonably well matched between the AKARI 9 μm and the Spitzer 8 μm channels (1σ scatter is 27%), keeping in mind not only filter variation but intrinsic stellar variability.

In contrast, the FIS catalog seems more often to not recover point sources seen in the Spitzer images. Without the AKARI images to compare, it is difficult to understand many of the FIS catalog detections in this confused and complex region. Only about 20 (<1%) of the FIS catalog detections correspond by eye to 70 μm Spitzer point or point-like sources. A significant fraction of the sources are clearly confused by structure in the nebulosity (e.g., there are "chains" of sources along emission ridges).

NGC 6997 has been identified historically as being a cluster superimposed either in front of or behind the NAN complex. Note that this cluster has been commonly confused with NGC 6996; see Laugalys et al. (2006b) and Corwin (2004). NGC 6997 has been recently studied by three papers in the literature. Zdanavičius & Straižys (1990) identified 15 cluster members, placing the distance at 620 pc (whereas the NAN is taken in that work to be at 550 pc) and the age at ∼10⁸ years. Villanova et al. (2004) double the cluster spatial size, finding ∼50 new probable or likely members, and placing the cluster at ∼760 pc and ∼350 Myr, in front of their distance for the NAN complex (which they take to be 1 kpc). Laugalys et al. (2006b) find ∼50 more members, place the cluster at 780–790 pc (now again behind the NAN), and estimate the age at 3 × 10⁸ years. It is identified in Figure 16 above.

For our discussion here, we do not attempt to redefine the cluster boundaries; we simply accept as legitimate members the 112 objects identified in these papers (and detected in our Spitzer survey) as cluster members, and it does not matter if the cluster is in front of or behind the NAN. On the whole, the stellar population of NGC 6997 indeed looks different than the rest of the cloud. None of the stars currently identified as NGC 6997 members have IRAC excesses, consistent with the cluster being older than (and probably unrelated to) the NAN complex. Twelve of the 112 objects are seen at 24 μm, and none at 70 μm. Out of these 12 objects, four are identified as high-confidence YSO candidates. The SEDs for the four high-confidence YSO candidates appear in Figure 28; they are GSC 03179-00959 (205622.0+443537.2), TYC 3179-1180-1 (205624.6+443921.3), TYC 3179-87-1 (205633.1+443810.3), and GSC 03179-00726 (205639.5+443821.6). There are two more lower-confidence YSO candidates (GSC 03179-01093 = 205633.3+443622.3 and GSC 03179-00980 = 205637.1+444009.2). These latter two objects have much lower-significance excesses at 24 μm. All of these YSO candidates appear in the complete tables of YSO candidates above.

Zoom In Zoom Out Reset image size

If these stars are truly members of NGC 6997, then these observed excesses could be only apparent excesses, due to source confusion or an unresolved lower-mass companion, or they could be real excesses. If they are real, then they are likely debris disks. The presence of such debris disks is broadly consistent with the age of the cluster being ∼350 Myr. Cieza et al. (2008) find, for the Hyades (∼650 Myr), that none of the FGK stars have debris disks, but that 2 of 11 A stars have debris disks. For these four high-confidence YSO candidates in NGC 6997, the spectral types from the literature are G7 V, B5 V, A3 V, and A7 III:, respectively; again, this is broadly consistent with the literature. A more complete study of the membership of this cluster would allow a more complete estimate of the disk fraction of this cluster.

We conclude that this cluster (as defined in the literature) is sufficiently different in mid-infrared properties from the NAN that it is unlikely to be related to the NAN, or at least much older than the rest of the complex.

Collinder 428 has been identified historically as being a cluster on the far east side of the complex that is likely unrelated to the NAN complex. Laugalys et al. (2007) concluded that the dust cloud in the direction of Collinder 428 is at the same distance as the NAN complex and that the star grouping is not a real cluster but rather a relatively transparent window in the dust cloud, and therefore not a separate object.

The stars in this region are on the edge of our map, and none of them are identified as YSO candidates using either IRAC or MIPS. Such a low disk fraction implies that if Collinder 428 is a real cluster, it is not part of the NAN complex. This is consistent with the Laugalys et al. (2007) interpretation that the stellar grouping is illusory due to a window in the dust cloud and not a separate entity.

Cambrésy et al. (2002) identified eight clusters based on 2MASS data. Clusters 1, 2, and 3a/b are part of the Gulf of Mexico, and cluster 5 is part of the Pelican cluster, as described above. We now discuss the remaining clusters. We do not define for further consideration any additional clusters associated with any of these objects, for reasons that will become apparent.

Cluster 4 appears in the 24 μm image as wispy faint (∼40 MJy sr⁻¹) nebulosity, with the nearest very faint point source 2' away (0.3 pc at 520 pc), and the nearest point source of a brightness range comparable to the clusters above nearly ∼75 away (1.3 pc).

Cluster 6 appears in the 24 μm image as the heart of a roughly arc-shaped patch of bright (≲650 MJy sr⁻¹) feathery nebulosity roughly 11' × 3' (1.7 × 0.5 pc at 520 pc). No point sources can be discerned on top of the nebulosity, though some faint point sources are seen close to the nebulosity.

Cluster 7 appears in the 24 μm image as a circular patch of bright (250–650 MJy sr⁻¹) nebulosity ∼4' (0.6 pc at 520 pc) in diameter. No point sources can be detected on top of this nebulosity.

Cluster 8, seen in 2MASS images as a tight, red clumping, appears as a single very bright point source at 24 μm, with no particular higher density of surrounding point sources at 24 μm.

In Table 8, we list the Spitzer fluxes (and cross-identifications) for those objects identified using the G09 methodology but not rediscovered using the MIPS-based selection mechanisms here.

In Table 9, we list the Spitzer fluxes (and cross-identifications) for those objects identified by the literature as YSO candidates but not rediscovered using either the MIPS-based selection mechanism that was the primary goal of this paper, or the IRAC-based selection mechanism from G09 and reported in the prior table. These objects may or may not be legitimate NAN members; we cannot determine this from our data set.

Note that six objects were identified in Witham et al. (2008) as bright in Hα in this region, but were not identified by us here as having an IR excess. All six of these objects are undetected at MIPS bands and on the edges of the IRAC maps such that they are only detected at two of the four IRAC bands. As such, they would not have been identified by our YSO identification approaches. They are objects 204654.78+434806.6, 204703.24+434930.0, 204704.82+ 434911.2, 205934.89+452355.9, 205938.69+452414.5, and 210029.50+443655.4.