THE SPITZER SURVEY OF INTERSTELLAR CLOUDS IN THE GOULD BELT. II. THE CEPHEUS FLARE OBSERVED WITH IRAC AND MIPS

Two epochs separated by 5–6 hr were used to take two complete sets of observations of the new SGBS clouds. The redundancy in coverage allowed for the rejection of transient phenomena (including foreground asteroids), the recovery of coverage lost by the blanking of bad pixels, and the full mapping at MIPS 70 μm where half the array was not working (each epoch was offset by half the array width). The unique AOR numbers for each epoch's observations are listed in Columns 3 and 4 for IRAC and columns 7 and 8 for MIPS. For the archival observations, only one epoch was obtained. Columns 5 and 9 list the respective dates when the IRAC and MIPS observations were taken. Likewise, Columns 6 and 10 list the version of the Spitzer Science Center (SSC) data pipeline that had been used to process the data prior to their being downloaded from the archive.

The total area mapped was 2.11 deg² with IRAC and 8.33 deg² with MIPS. MIPS mapped a larger area than IRAC because its minimum scan lengths were longer than the diameter of the compact regions of high extinction. The regions for which we have four-band IRAC fluxes are slightly smaller than the areas listed in Table 1. This is because the IRAC detectors are paired into two different pointing offsets resulting in slightly different coverage patterns between 3.6/5.8 μm and 4.5/8.0 μm. The IRAC observations have a total integration time of 48 s per point split between the two epochs. Each epoch consisted of two dithers. The SGBS MIPS observations were taken with the fast-scanning mode using a 240 arcsec step size and a cross-scan overlap. This gave a total per point integration time of 31.4 s at 24 and 70 μm and 6.1 s at 160 μm.

The basic calibrated data (BCD) were downloaded from the Spitzer archive and then ingested into the c2d/SGBS analysis framework (for a full description, see Evans et al. 2007 and Huard et al. 2009, or separately Harvey et al. 2006 for IRAC, and Rebull et al. 2007 for MIPS). In brief, the data were inspected and custom data masks were created to identify bad pixels. The data were then corrected for instrumental effects including bleed over, saturation, and banding effects for IRAC and jailbar and stim flash latents for MIPS. The improved data were mosaiced with the MOPEX package (Makovoz et al. 2006) and source extraction with the c2dphot tool was performed independently at each wavelength and epoch (Evans et al. 2007; Huard et al. 2009). The c2dphot tool is a derivative of the DAOPHOT source extraction and photometry tool (Schechter et al. 1993).

IRAC and MIPS 24 μm photometry was performed with c2dphot during source extraction. Each source was characterized by a point-spread function (PSF) fit performed in a 9 pixel (3.6–5.8 μm) or 11 pixel (8 μm and 24 μm) wide box, the local background level was one of the free parameters. Aperture fluxes were measured in a 7 pixel wide box centered on the source and flux uncertainties were estimated from the goodness of the PSF fit. The pixel size of the IRAC mosaics is 1.2 arcsec, and the pixel size of the MIPS 24 μm mosaics is 2.6 arcsec. The MOPEX point-source fitting package was used for the 70 μm photometry (Makovoz et al. 2006). These methods are discussed in detail in the c2d and SGBS delivery documentation (Evans et al. 2007; Huard et al. 2009). The separate epoch/wavelength source lists were band-merged together with the Two Micron All Sky Survey (2MASS) catalog (Skrutskie et al. 2006) and cross-identifications were made at better than 2 arcsec accuracy. Each source was then characterized spectrally and for the quality of detection. The catalog was "band-filled" to produce upper limit flux estimates for sources that were not detected at all wavelengths (Evans et al. 2007; Huard et al. 2009).

Table 3 lists the original survey/instrument, filter, nominal wavelength, flux zero-point F_0pt, and limiting magnitude of the ten bands from the band-merged catalog (2MASS J–K_s; IRAC 3.6–8.0 μm and MIPS 24–160 μm) plus additional bands from the second HST Guide Star Catalog (GSC-II, B–I), the IRAS point-source catalogs (12–100 μm), and the Submillimeter Common User Bolometer Array (SCUBA) data archive (450–850 μm). Data for these additional bands were only added for YSO candidates, as described in Section 5.1. The photometric system is based on the Vega magnitude system using the flux zero points taken from the c2d delivery documentation (Evans et al. 2007). These are within 1%–2% of the SSC's IRAC¹¹ (Reach et al. 2005) and MIPS¹² zero points. We use the c2d zero points to maintain compatibility with the c2d delivery documentation (Evans et al. 2007). The minimum uncertainties for individual flux measurements are 4% and 8% for IRAC and MIPS (24 μm and 70 μm), respectively. These uncertainties do not include the zero-point/absolute calibration uncertainties of 1.5% for IRAC, 4% for MIPS 24 μm, 20% for MIPS 70 μm (Evans et al. 2007). The absolute Spitzer calibration uncertainties are lower than the individual measurement uncertainties by a factor of 2 or more, except for the MIPS 70 μm band, which has a significantly higher calibration uncertainty. The limiting magnitudes are identified in Section 3.4 as the turnover in the source count distributions.

Table 3. Spectral Bands in This Survey

Origin	Filter	Wavelength (μm)	F0pt (Jy)	Limiting Magnitudea	θb	Ref
GSC-IIc	B	0.44	4260.0	...	17	1, 8
GSC-IIc	V	0.55	3640.0	...	17	1, 8
GSC-IIc	R	0.64	3080.0	...	17	1, 8
GSC-IIc	I	0.79	2550.0	...	17	1, 8
2MASS	J	1.24	1594.0	16.3	25	2
2MASS	H	1.66	1024.0	15.5	25	2
2MASS	Ks	2.16	666.7	15.3	25	2
IRAC	IRAC1	3.6	280.9	17.2	17	3, 6
IRAC	IRAC2	4.5	179.7	16.7	17	3, 6
IRAC	IRAC3	5.8	115.0	15.5	19	3, 6
IRAC	IRAC4	8.0	64.13	14.3	20	3, 6
IRAS	IRAS1	12.0	...	...	45	4
MIPS	MIPS1	24.0	7.14	10.2	60	3, 7
IRAS	IRAS2	25.0	...	...	46	4
IRAS	IRAS3	60.0	...	...	47	4
MIPS	MIPS2	70.0	...	...	18''	3, 7
IRAS	IRAS4	100.0	...	...	50	4
MIPS	MIPS3	160.0	...	...	40''	3, 7
SCUBA	SHORT	450.0	...	...	75	5
SCUBA	LONG	850.0	...	...	14''	5

Notes. ^aTaken as the turnover in the S/N limited source count distribution. ^bThe effective angular resolution at each wavelength. See quoted references for details. ^cThe GSC-II data have been converted into the Johnson–Cousins photometric system from the natural systems used by its component surveys References. (1) Bessell 1979; (2) Skrutskie et al. 2006; (3) Evans et al. 2007; (4) Neugebauer et al. 1984; (5) Holland et al. 1999; (6) Fazio et al. 2004; (7) Rieke et al. 2004; (8) Lasker et al. 2008.

Download table as:  ASCIITypeset image

The c2d data used in this paper come from their final data release¹³ (DR4) and use the same target selection and observation strategy as the SGBS program. A significant difference was that the c2d cores data were taken using the MIPS small map photometry mode rather than the fast-scanning technique. As a result, the 70 μm coverage on the c2d cores can actually be smaller than the IRAC area (Evans et al. 2007). No 160 μm data were taken for the c2d isolated cores (requesting 160 μm data imposes a scheduling limitation as it can only be taken in particular "cold" campaigns). The L1148+L1152+L1155 region contains three c2d cores in close proximity so the entire area was covered by a single SGBS MIPS fast scan to obtain a complete 160 μm coverage. The only regions for which 160 μm data were not taken were L1221 and L1251.

At the time of writing, the c2d studies on the individual first list cores have already been published or are in the process of being published. L1148 was studied by Kauffmann et al. (2005), who discovered a new very low luminosity (VeLLO) Class 0 YSO. L1155C, L1152, and L1228N were studied by Chapman & Mundy (2009) to examine the extinction law from 3.6 to 24 μm. Although L1228N was observed by c2d, we use the First Look Galactic data set as it covers both L1228N and L1228S. An analysis of the small YSO group in L1228S was published by Padgett et al. (2004). L1221 was extensively studied by Young et al. (2009), including follow-up observations with the VLA. L1251B was studied by Lee et al. (2006), who compared their YSO detections with continuum submillimeter data. This was then followed up by a molecular line study (Lee et al. 2007). The data set was built with the aim of maximizing the coverage with a (relatively) uniform sensitivity. We chose not to add deeper data (e.g., the PID #3656 data used by Chapman & Mundy (2009)) as we wish to have all regions covered to approximately the same depth.

The L1152, L1155, and L1221 studies used data from the DR4 c2d data release (Chapman & Mundy 2009; Young et al. 2009), the same one used by this paper. The L1251 study, however, used data from an earlier data release (DR3) that were processed through version S11 of the SSC pipeline (Lee et al. 2006). Likewise, the First Look study of L1228 used data from version S9.1 of the SSC pipeline (Padgett et al. 2004). The SSC pipeline is constantly being revised to improve calibration and reduce instrumental artifacts. In particular, pipeline versions up to S9.1 used a preliminary flux calibration model.¹⁴ Improvements in the calibration will change the photometry for regions where a newer pipeline has been used than that in the original study. This is most significant for the L1228 region where it was found that fluxes measured from S14/16 data had increased by an average of 16% against the S9 version of the data. The greatest increase was at 70 μm, which increased by almost 40%.

Figure 2 shows RGB color composites of 4.5 μm (blue), 8.0 μm (green), and 24 μm (red) toward the regions observed for the SGBS. The green haze that covers a large area of the images comes from the 8.0 μm background, which is believed to arise predominantly from 7.7 μm polycyclic aromatic hydrocarbon (PAH) emission (Flagey et al. 2006). In L1241, the haze appears along the edge of the image and is anticorrelated with regions of higher 160 μm emission (see Section 6). While the mosaic processing removed a number of instrumental effects, it was unable to correct for the bleeding at 8 μm that appears in the color composite images as green smearing upward from the bright center of NGC 7023. The red point source in the L1148 map is the heavily reddened L1148-IRS source studied by Kauffmann et al. (2005).

Zoom In Zoom Out Reset image size

The brightest extended structure at all wavelengths in this survey is the NGC 7023 reflection nebula. Figure 3 shows three-color composite images covering the visible (Digitized Sky Survey-II; left image), infrared (IRAC; middle image), and far-IR/submillimeter wavelengths (MIPS; right image). The extended cold dust emission, seen in silhouette in the first panel and emission in the last panel, forms a faint cross with HD 200775 at its intersection. The vertical axis of the cross is formed by the north-south filament that includes L1172 and L1174. The east-west axis is formed by the reflection nebula forming in this filament. The outflow from HD 200775 is currently inactive, but it has generated an asymmetric east-west biconical cavity that is filled with hot atomic gas (Fuente et al. 1998). The larger western lobe of this cavity appears as a hole in the left-hand panel through which background stars are clearly visible. The eastern lobe is truncated by a dense material that is being heated and photodissociated by the YSO (Fuente et al. 1998).

Zoom In Zoom Out Reset image size

The reflection nebula is confined to the north by a dense filament. This same structure is shown in the strong 3.6 μm emission on the leading edge of a colder dense material revealed by the 850 μm dust emission (see the Appendix). To the south of NGC 7023 is the L1172D dark nebula that appears in silhouette in the visible image, but reveals three embedded sources in the mid-infrared image. The dust itself emits at the longest wavelengths and shows the bright point source associated with the L1172 SMM 1 embedded protostar. The extended 8 μm emission (the red haze in the central image) is anticorrelated, however, with the high visual extinction in the first image and the dense gas in the third image.

Tables 4 and 5 list the number of objects detected in various combinations of the IRAC or MIPS bands within the SGBS Cepheus Delivery Catalog, which contains a subset of the data included in this paper. Specifically, it includes the five regions: L1148, L1172, L1228, L1241, and L1247+L1251. The other regions (L1155, L1152, and L1221) are available from the SSC as c2d data products. The source detection for SGBS data is defined as an object with a peak intensity ⩾3 × the local rms in its respective band. A 2MASS detection constitutes an object seen at ⩾10 × the local rms in both the H and K_s bands (a design constraint of the 2MASS catalog itself; see Skrutskie et al. 2006).

Table 4 shows that a total of 71,085 sources were detected in at least one IRAC band, and of those, 6518 were detected in all four bands. Table 5 lists the number of source detections for each of the four IRAC bands and the first two MIPS bands for several different signal-to-noise thresholds. It shows that the majority of these sources were predominantly detected at shorter wavelengths. An important distinction between Tables 4 and 5 is that the former only refers to the area common to all Spitzer wavelengths, whereas the latter lists statistics for the entire region of the delivery catalog.

Figure 4 shows source count histograms of the number density of sources per magnitude interval per square degree for six bands from the delivery catalog. The gray curve shows all sources. The limiting magnitudes listed in Column 5 of Table 3 are taken as the magnitudes where the black curves in Figure 4 turn over. The total source count histogram at 3.6 μm is double-peaked. The 4.5 μm source count histogram shows a second peak, but the effect is far less pronounced. Double peaks were also found in the source count histograms for Perseus (Jørgensen et al. 2006) and Chamaeleon II (Porras et al. 2007). Jørgensen et al. (2006) showed that the double peak was not due to the presence of background galaxies by comparing counts of on-cloud and off-cloud sources and that it must derive from something, like extinction, that is common to the entire region. The solid black curves in Figure 4 plot only those sources that meet our detection criteria. The peaks of the black curves for 3.6 μm (IRAC1) and 4.5 μm (IRAC2) are coincident with the first of the double peaks.

Zoom In Zoom Out Reset image size

Overplotted in Figure 4 are the source count histograms of Galactic infrared sources toward Cepheus estimated from the model of Wainscoat et al. (1992) updated by J. Carpenter (2001, private communication) to apply to the Spitzer bands. The source counts agree with the model at low magnitudes before diverging at high magnitudes. The divergence from the model increases with increasing wavelength. An excess of sources above the model can be attributed to the detection of extragalactic sources that are not present in the Wainscoat model. The majority of the 24 μm (MIPS1) sources are in excess of the Wainscoat model and are therefore probably extragalactic in origin.

Young stellar objects (YSOs) and background galaxies can have similar infrared colors, but they can be differentiated by the fainter apparent magnitudes of the galaxies. There are several published schemes for identifying YSO candidates from their Spitzer colors, each relying on the specific subsets of the Spitzer wavelength coverage and the equivalent 2MASS fluxes (e.g., Harvey et al. 2006, 2007; Rebull et al. 2007; Allen et al. 2007; Gutermuth et al. 2008). These schemes are commonly calibrated against a known galaxy catalog. SWIRE (the Spitzer Wide-area Infrared Extragalactic Survey, another of the Spitzer legacy programs) was specifically designed to observe the infrared extragalactic background along sightlines that avoided as much foreground (Galactic) contamination as possible (Lonsdale et al. 2003). Thus, it provides an ideal calibration data set to use with any galaxy-rejecting protocol.

IRAC observations were targeted toward regions of high visual extinction with the expectation that these regions will harbor the highest concentration of YSOs. A side effect of the different operational modes of IRAC and MIPS is that the MIPS data cover an area four times as large as the IRAC data. A galaxy-rejection scheme that used IRAC-only data would be unusable across 3/4 of our surveyed area. Therefore, we adopt different schemes for the common (IRAC+MIPS) and the MIPS-only areas. A significant factor in both schemes is a 24 μm MIPS detection. Since that band is not as sensitive as the IRAC bands, a third scheme is adopted for sources detected only with IRAC.

Table 6 lists the YSO candidates identified toward the regions in our survey. The first three sections of Table 6 list the YSO candidates identified by the 5-Band, IRAC-only, and 2MASS/MIPS schemes, respectively (described below). The first column lists the source number. The second column lists the Spitzer catalog position identifier of each source (a contraction of the source's seven-digit right ascension and declination without delimiters). All positions are given in J2000. Column 3 lists cross-identification names from the literature. Columns 4–9 list the flux and statistical errors or where appropriate the band-filled upper limit. Column 10 lists a series of flags denoting which candidate identification methods identify the source. The letters F, I, and M are shown for sources identified by the 5-Band, IRAC, and 2MASS/MIPS methods, respectively. A source may satisfy the rule sets of more than one method, but it is assigned to a group based on the order F > I > M. This catalog includes data previously published in a series of c2d papers examining individual dark clouds. Column 11 lists references to these and other studies.

4.1.1. 5-Band Identifications

Harvey et al. (2007) used a 5-Band (4.5, 8.0, and 24 μm, plus 2MASS H and K_s) scheme to estimate the probability that a given source is a background galaxy. The unnormalized probability P_gal is the product of a series of probabilities that a given source is a background galaxy. Each of these probabilities is based on the selection criteria from a different color–magnitude diagram. The final value of P_gal is moderated by a series of additional factors that include whether the source is extended at 3.6 μm or 4.5 μm (see Table 1 of Harvey et al. 2007 for a full list). Additionally, the scheme requires that a source be detected at each IRAC band and at MIPS 24 μm, irrespective of whether that band was actually used to construct P_gal.

Under the 5-Band scheme, a catalog is filtered to reject those sources that can be adequately modeled by a stellar photosphere. P_gal is calculated for the remaining sources and these are then filtered to retain sources with a suitably low value of P_gal, i.e., rejecting sources that are statistically likely to be background galaxies. From their study of Serpens, Harvey et al. (2007) found that an upper limit of log(P_gal) < −1.47 rejected the galaxies from their SWIRE control catalog. This method creates a catalog of YSO candidates that is largely free from background galaxies. The catalog will be luminosity limited partially because it requires [24] > 10 (fainter objects at [24] are preferentially background galaxies, but this cutoff will also exclude faint YSOs).

Figure 5 shows a histogram of log(P_gal) for sources in the Cepheus Catalog with detections in all IRAC bands and MIPS 24 μm. The peak at −5 contains the majority of the YSO candidates identified by this method. There is also a small tail of sources to the left of the −1.47 divide (the vertical dashed line). There are 12 sources in the tail with −2.5 < log(P_gal) < −1.47 and three (25%) of them are previously known YSO candidates. We choose to retain the canonical dividing line for YSO candidates for consistency with the c2d studies (Harvey et al. 2007; Evans et al. 2009) and because moving the division by any significant amount would exclude the three known sources. If half of the tail sources were background galaxies, it would represent a contamination of ∼5% in the total number of YSO candidates identified by the 5-Band method.

Zoom In Zoom Out Reset image size

Figure 6 shows a sequence of color–color plots for sources detected in all four IRAC bands and shows the cutoffs Harvey et al. (2007) used to construct P_gal. Dashed lines show "fuzzy" limits, while solid lines show hard limits. The left panel in each pair shows the SGBS data. Black filled circles and crosses are, respectively, the point-like and extended sources identified as YSO candidates by P_gal, while dark gray circles are sources identified as point-like (filled circles) and extended (open circles) galaxies. Also shown by pale gray points are sources that were identified as stellar photospheres via their SEDs. The right-hand panel of each pair of plots in Figure 6 shows a contour plot of the SWIRE Catalog after it has been processed in a manner similar to the SGBS Catalog. It can be seen how the limits used in the calculation of P_gal have been chosen to reject regions of parameter space where there are significant numbers of SWIRE galaxies. These plots show that the galaxies are tightly clustered and that there is no significant overlap between the different candidate types. The 5-Band scheme identified 98 YSO candidates and one possible YSO candidate with a band-filled (upper limit) flux at 8 μm (shown as an open square in Figure 6). The sources that satisfy the 5-Band scheme are listed with an F flag in Column 10 of Table 6.

Zoom In Zoom Out Reset image size

4.1.2. IRAC-only Identifications

For those sources that have no detection at 24 μm, we must use the selection rules based on IRAC colors alone. For this, we use the selection rules from Harvey et al. (2008). They define ([4.5]–[8.0] < 0.5 and [8.0]>13 − ([4.5]–[8.0])) to reject galaxies and stars. The upper row of Figure 7 shows a plot of [4.5]–[8.0] versus [8.0] for sources without an MIPS 24 μm detection and comparative plots of the 5-Band and SWIRE sources. Of the YSOs identified by the 5-Band scheme, 12 would be misidentified by the IRAC-only scheme as stars due to their small [4.5]–[8.0] color and six bright galaxies would have crept into the sample.

Zoom In Zoom Out Reset image size

The sources that satisfy the IRAC-only scheme are listed with an I flag in Column 10 of Table 6. A source flagged with a lowercase f has a 24 μm upper limit flux, but would otherwise satisfy the 5-Band scheme. A comparison of sources with I and f flags shows that the IRAC-only scheme and the 5-Band scheme are consistent for sources without a 24 μm detection. A total of 10 additional YSO candidates were identified by the IRAC-only scheme.

An additional constraint can be imposed in color–space as external galaxies are often rich in PAH emission compared with YSO sources. Gutermuth et al. (2008) proposed using the [4.5]–[8.0] versus [3.6]–[5.8] and [5.8]–[8.0] versus [4.5]–[5.8] color–color diagrams to reject background galaxies with strong PAH emission. The sources that match these rules are shown with triangle rather than circle markers in Figure 7. No IRAC YSO candidates have evidence of PAH emission from this technique, but two 5-Band YSO candidates are flagged. One of these is L1148-IRS.

4.1.3. 2MASS/MIPS Identifications

An inherent strength of the 5-Band scheme is that it relies on a broad range of color combinations, but it can only be used in regions where there is IRAC coverage. Rebull et al. (2007) proposed a scheme to identify YSOs using MIPS 24 μm and 2MASS K_s and the selection rules K_s − [24]>2 and K_s < 14. To this we add the limit [24] < 10 for consistency with the 5-Band scheme. The lower row of Figure 7 shows color–magnitude plots of K_s − [24] versus K_s for sources without an IRAC detection and comparative plots of the 5-Band and SWIRE sources. Column 10 of Table 6 includes an M flag for the sources that would have been identified as a YSO candidate by this method. A significant fraction of 2MASS/MIPS YSO candidates were found within 1 magnitude of the selection limits. These are marked as open circles in Figure 7 and listed with a lowercase m in column 10.

The 2MASS/MIPS scheme would have misidentified 19 of the 5-Band YSO candidates as background galaxies and stars. From Figure 7, we can see that the majority of these appear in the region of background galaxies with a K_s − [24] color of ∼10. This is approximately equivalent to the spectral index of an embedded protostellar source (a Class I YSO, see the next section and Figure 8). This shows a further advantage of the 5-Band scheme in that it is more sensitive to embedded sources as it does not rely heavily on K_s. Based on the MIPS scheme, 4–5 galaxies could have crept into the YSO sample. A total of 24 additional YSO candidates were identified by the 2MASS/MIPS scheme that were not identified by the 5-Band and IRAC-Only schemes.

Zoom In Zoom Out Reset image size

4.1.4. The Final Catalog

YSOs can be separated into a series of four evolutionary classes depending on either their infrared spectral index (Wilking et al. 1989) or the mean frequency of their SED (Myers & Ladd 1993). The spectral index classification of a source as a Class II or Class III YSO is generally correlated with its respective classification as a Classical T Tauri (CTTS) or Weak-Line T Tauri (WTTS) star (Wilking et al. 1989; André & Montmerle 1994). Class 0 and Class I protostars are younger embedded YSOs and are differentiated from each other by the amount of material remaining in their envelopes. The envelope appears as a colder component to the SED and peaks toward the submillimeter. The original definition of a Class 0 protostar was an object whose submillimeter luminosity (λ> 350 μm) contributed greater than 0.5% to the source's total bolometric luminosity, which is equivalent to a protostar-to-envelope mass ratio of less than one (André et al. 1993). The second method of YSO classification uses a bolometric temperature calculated for a blackbody that has the same mean frequency as the source's SED. The value of the bolometric temperature decreases from Class III to 0 as the SED is increasingly dominated by long wavelength emission. The results for this method are described in Section 5.2.

Table 7 lists the derived properties of the YSOs from Table 6. The first column lists the index number of the YSO. Column 2 lists the dark cloud from Figure 1 with which the YSO candidate is associated. Locations that have a visual extinction less than one magnitude in the Dobashi et al. (2005) maps or are not part of an identifiable YSO group (see Section 6.3) are listed as "off-cloud." Column 3 lists the spectral index of the source, α_IR, as given by

$$\begin{equation} \alpha _{\rm IR} = \frac{d\log (\lambda F_{\lambda })}{ d\log \lambda }, \end{equation} \tag{ 1 }$$

where F_λ is the monochromatic flux density at wavelength λ. The index was calculated by a least-squares fit to all available data in the range 2MASS K_s (2.2 μm) to MIPS 24 μm.

The mapping between the YSO evolutionary sequence and the spectral index proposed by Wilking et al. (1989) and amended by Greene et al. (1994) and André & Montmerle (1994) is Class I (α_λ ⩽ 0.3), "flat" (0.3>α_λ ⩾ −0.3), Class II (−0.3>α_λ ⩾ −1.6), and Class III (−1.6>α_λ). The decreasing spectral index is attributed to the decline in the amount of circumstellar dust with advancing YSO evolution. The associated class for each source is listed in Column 4 of Table 7. "Flat" spectrum sources have spectra intermediate between embedded and T Tauri stages. Class 0 YSOs cannot be distinguished from Class I YSOs via this method as it is not sensitive to the long wavelength part of the SED where the majority of the envelope emission radiates (Enoch et al. 2009). Radiative transfer modeling of YSO SEDs has shown that the amount of dust observed around a YSO, and thus its spectral index, will depend on the angle of inclination of the circumstellar disk to the observer (Robitaille et al. 2006; Crapsi et al. 2008). This degeneracy could possibly be broken by using longer wavelength data (see Section 5.2 for a comparison between the number of YSOs classified by each method).

The majority of the data points used to calculate α_IR are in the IRAC regime. For those YSO candidates without an IRAC detection, α_IR is only calculated from 2MASS K_s and MIPS 24 μm—effectively the K_s-[24] color. To test the robustness of using just the 2MASS/MIPS color, the top of Figure 8 shows K_s-[24] versus α_IR for the YSO candidates detected with IRAC. The tight correlation follows a linear relationship, thus proving that the equivalent color can be an acceptable proxy for a spectral index calculated using more intermediate points.

The solid line through the points in Figure 8 shows the theoretical relationship between K_s-[24.0] and α_λ. For two generic bands, A and B, the color and the spectral index are linearly related such that m_A − m_B = C₁α_λ + C₂, where C₁ = −2.5X, C₂ = −2.5 + 2.5Y, X = log(λ_A) − log(λ_B), and Y = log(F_0,A) − log(F_0,B). The wavelength of each band is λ_A and λ_B, the flux zero points are F_0,A and F_0,B, and the color is (m_A − m_B). Using the data from Table 3, the coefficients for K_s-[24.0] are C₁ = 2.63 and C₂ = 7.55. A linear regression to the data, shown by the dashed line, gives coefficients of C₁ = 2.64 ± 0.06 and C₂ = 7.73 ± 0.06.

The color-α_IR equation can be used to recompute the class boundaries in terms of a source's K_s-[24.0] color such that they become K_s-[24.0] = 8.32, 6.74, and 3.32 for the boundaries between Class I/"flat," "flat"/Class II, and Class II/III, respectively. These values agree with those used by Lee et al. (2006) and Rebull et al. (2007). These recomputed boundaries are shown in Figure 8 by the dotted lines.

The lower panel of Figure 8 shows a histogram of the YSO spectral indices. The black line histogram shows sources with IRAC coverage. The range of values of α_λ is −2.28 to 1.63. The gray histogram shows all sources and includes 2MASS/MIPS YSOs. Both distributions peak around −1 and show an increase in the number of sources from Class I to Class II. The peak of the Cepheus spectral index histograms matches the peak of the distributions seen in Serpens (Harvey et al. 2007), Lupus (Merín et al. 2008), and IC 5146 (Harvey et al. 2008). The range of extreme spectral indices is less than that in IC 5146 (Harvey et al. 2008).

The number of sources classified in each class via α_λ and that number as a percentage of the Spitzer-identified YSOs is listed in Column 2 of Table 8. The row for Class 0 sources is left blank for α_IR, as they can only be distinguished by their bolometeric temperature. We can compare the relative fraction of α_IR classes for the Cepheus Flare with the values for other regions from the c2d survey (Figure 5 of Evans et al. 2009). The percentage of sources in each class closely matches the numbers for the entire c2d survey (shown in Column 4 of Table 8). Star formation in the Cepheus Flare is comprised of a series of small YSO groups, isolated cores, and a single (relatively) large YSO group (see Section 6). That this mixture gives the Cepheus Flare a relative number of sources equal to the c2d survey may just be a coincidence or it could be that the mixture of star-formation modes in Cepheus mirrors the balance of modes across the wider c2d sample.

Of the individual c2d regions, Cepheus most closely matches the α_IR class profile of Serpens (Harvey et al. 2007), which is also very similar to that for the full c2d survey (Evans et al. 2009). Cepheus also has the same relative number of Class I sources as Serpens (Harvey et al. 2007). The relative number of Class I sources in Cepheus is ∼10 percentage points higher than that seen in the Chamaeleon II and Lupus clouds although it is ∼5 percentage points lower than that seen in the IC 5146 and Perseus clouds (Harvey et al. 2008; Evans et al. 2009).

Kun et al. (2008) include in their review of the Cepheus star-formation region a list of Classical and Weak-Lined T Tauri stars whose pre-main-sequence nature has been confirmed by spectroscopic observations. There are 31 Spitzer-identified YSOs from our sample that appear in that list (these have a reference to the Kun review in the last column of Table 6). Twenty-five of these sources (80%) are classified as Class II by their α_IR value. This match agrees with the usual interpretation of a YSO with a Class II infrared spectrum as most probably being a T Tauri protostar.

There are four sources on the Kun et al. list that are among the previously known YSOs listed in Table 16 that were coincident with our mapped area, but which were not identified as YSO candidates by the Spitzer schemes. Despite not being identified as YSOs, three of these four sources were coincident with entries in the Spitzer Cepheus catalog. Therefore, 34 out of 35 of the Kun et al. stars had infrared emission that was detectable by Spitzer. It should be noted, however, that the Kun et al. T Tauri list is from a project that is still underway and is not necessarily complete.

Our Class III to Class II ratio is 1/8, whereas WTTS dominates CTTS in spatially complete surveys. For example, the WTTS/CTTS ratio in Taurus is 8 (Neuhaeuser et al. 1995) and 3/2 in IC 348 (Luhman et al. 2003). Infrared selection schemes will naturally be less sensitive toward objects with smaller infrared excesses, i.e., Class III YSOs. The estimates of completeness can only be made if there is a previous census of WTTS to compare against. In a comparable study of the Lupus III region, Merín et al. (2008) estimated their completeness for Class III sources with infrared excesses against objects with no infrared excess as ∼50%. If the low rate of T Tauri sources in the Cepheus cores without detectable infrared excesses is real and similar to Lupus III, it would further reinforce the idea that the YSO population is comparatively young.

Figure 7 shows that no 2MASS/MIPS YSO candidates are detected with a K_s-[24] color greater than ∼5 (or 6). From Figure 8, we see that this means that the 2MASS/MIPS method is significantly biased toward the detection of Class II and Class III YSOs. Deeper K_s data are needed to make a more complete survey of YSOs in the MIPS-only regions.

Figure 9 shows color–color diagrams of the 143 Cepheus YSOs listed in Table 6. A source is included in a given plot if it has been detected at each of the wavelengths shown in that plot. Red markers show Class I sources, green markers show flat spectrum sources, blue markers show Class II sources, and purple markers show Class III sources. Robitaille et al. (2006) generated a database of 20,000 YSO models and used them to delineate a series of regions in the color–color space that correspond to three different phases of YSO evolution. They termed these evolutionary phases "Stages" to differentiate them from the equivalent "Classes" (a purely observational parameter that describes a source's infrared signature and may be influenced by reddening, e.g., as a result of the varying disk orientation). The approximate boundaries of the regions containing the different evolutionary states are shown in each of the three panels in Figure 9 (see Figure 23 of Robitaille et al. 2006).

Zoom In Zoom Out Reset image size

Allen et al. (2004) compared the theoretical Spitzer colors of YSOs with data from the four embedded clusters. The region they identified as corresponding to the approximate domain of Class II sources contains almost all of our Class II candidates as shown by the thick lined box in the first panel of Figure 9. The small six-sided region coincident with the Allen box shows the region where most of the Stage II Robitaille models lie. Our Class II sources cluster around this region, but there are a number of sources that lie just above it. These could be Class II sources that still retain some degree of reddening. The Class III sources are clustered around the zero point on each axis. Two large regions in the first panel are divided by a line that runs through the Stage II region and the Class II box. The region above this line contains the majority of the Robitaille Stage I models, whereas the region below it is where any Stage may be present. Our Class I sources agree quite well with the Stage I area.

The middle panel of Figure 9 shows IRAC and MIPS colors with the expected color space limits of the three Robitaille Stages (Stage III, II, and I running left to right). The points show the Cepheus YSO candidates with the same color code as the first panel. In this color space, there is also excellent agreement between the stage spaces and the classes. The green points show flat spectrum sources. These do not correspond to a specific stage, but are found along the Stage I/II boundary, as expected. Evans et al. (2009) found a similar result for the entire c2d survey and showed that the extinction correction of the individual fluxes only marginally improved the agreement between the classes and the stages. The stage regions take account of different angles of inclination, and this is partially why the Class I sources scatter over such a large area (Robitaille et al. 2006). The angle of inclination of a disk system to the observer is important as it can have a strong effect on the source's infrared signature and lead to the source being misclassified (Crapsi et al. 2008).

The last panel of Figure 9 shows the H − K_s versus J − H 2MASS colors of the Cepheus YSO candidates. The YSOs are not as separated as in the other two color spaces, but there is a trend for the Class I and Flat sources to scatter away from the main locus. The box around this locus shows the expected domain of reddened stellar photospheres from the Robitaille models, while the enclosed region to the right shows the region where any Stage evolutionary model can be present (Robitaille et al. 2006).

5.1.1. Guide Star Catalog

5.1.2. IRAS

5.1.3. Submillimeter

Figures 10, 11, 12 and 13 show plots of the SEDs of sources classified as Class I, Flat, Class II, and Class III, respectively. The open circles show, where available, the photometry from GSC-II, 2MASS, IRAC, MIPS, IRAS, and SCUBA. The arrows show the position of flux upper limits. The top left-hand corner of each plot is labeled with the index of the YSO. If the source index is followed by "x0.01," it indicates that the SED has been scaled downward by two dex in order to place it on the same grid as the other SEDs. In Figures 12 and 13, the gray lines show two comparison SEDs that have been normalized near the peak of the dereddened SED (usually the 2MASS J band). The solid gray line is a NEXTGEN profile for a K7 star (Hauschildt et al. 1999) and the dashed gray line with error bars is the median SED for a T Tauri star in Taurus (Hartmann et al. 2005). These SEDs are discussed in more detail in Section 5.4.

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

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For the Class I YSOs in Figure 10 with three or more detections longward of 65 μm, we fit simple graybody SEDs assuming β = 2 (see Kirk et al. 2007 for details) to characterize their submillimeter luminosities. Table 12 lists the derived parameters from these fits. Column 1 lists the YSO index, Column 2 lists the fitted dust temperature, and Column 3 lists the submillimeter luminosity integrated under the fitted graybody between 350 μm and 2000 μm.

An alternative method to the spectral index for deriving the evolutionary classification of a YSO is to estimate the source's bolometric temperature, T_bol, which is the temperature of a blackbody that has the same mean frequency as a source's observed SED (Myers & Ladd 1993). T_bol can classify embedded protostars more effectively than the spectral index as it uses the entire spectrum and not just the infrared portion used for the spectral index (Enoch et al. 2009). Column 5 of Table 7 lists T_bol for each protostar calculated using all available data points. A Class 0 source has T_bol < 70 K, Class I has T_bol = 70–650 K, Class II has T_bol = 650–2880 K, and Class III has T_bol > 2880 K (Chen et al. 1995). The T_bol boundaries for the equivalent of the intermediate flat spectrum class were not set by Chen et al. (1995), but Evans et al. (2009) recently suggested boundaries of 350 K and 950 K. Column 6 of Table 7 lists the class for each YSO as derived from the Chen et al. (1995) boundaries with the Evans et al. (2009) modification for flat spectrum sources.

The number of YSOs in each class as classified by T_bol is listed in column 3 of Table 8. The spectral index cannot classify a source as a Class 0 source, so no total is listed the α_IR column for such objects. While there is general agreement between the relative number of each class of YSOs between both schemes (i.e., a few Class III sources compared with many Class II sources), the bolometric method tends to skew a source's classification toward an earlier class (if the class changes at all). Out of the 87 sources classified as Class II by their value of α_IR, six (7%) were classified as Class 0 or I by their value of T_bol. For example, YSO #68 has a spectral index of −0.51 that classifies it as a Class II source, whereas it has a bolometric temperature of 38 K that places it firmly in the Class 0 regime. The spectral index is only calculated on flux data up to 24 μm, so flux data longward of 24 μm, as would be expected to dominate the SED of an embedded protostar, will not factor into a source's classification.

Conversely, out of the 62 sources classified as Class II by their value of T_bol, only YSO #45 had a T_bol Class earlier than Class II. As mentioned in Sections 4.2 and 4.3, the classification of a source by its α_IR is sensitive to the angle of inclination of the circumstellar disk to the observer. Radiative transfer models have shown that Class II sources viewed at high angles of inclination can have SEDs that appear Class I like (Robitaille et al. 2006; Crapsi et al. 2008). The low frequency of this effect between the α_IR and T_bol classifications for the Class II and Class I sources would suggest either that the inclination angle effect is not significant in this sample or that the T_bol classification is also effected by the inclination angle effect. In the latter case, the magnitude of the inclination angle effect could only be assessed by confirming the individual sources's evolutionary status with a spectrographic follow-up campaign.

Enoch et al. (2009) examined the effect on T_bol of excluding a 160 μm data point. They found that T_bol will be overestimated, and the ratio of Class 0 to Class I sources will be skewed toward Class I sources. For the Cepheus data, we found that the exclusion of all data longward of 150 μm (i.e., the 160 μm and submillimeter data listed in Table 11) did not significantly affect the numbers of sources classified as Flat or Class II and III. It did, however, affect the relative number of Class 0 and I sources. The number of Class 0 and I sources classified by T_bol calculated using the submillimeter data is 17 and 11, respectively (as shown in Table 8), whereas the number of Class 0 and I sources classified by T_bol calculated without the submillimeter data is 4 and 16, respectively, in agreement with the Enoch et al. (2009) finding.

We retain the submillimeter data points when calculating the value of T_bol quoted in Table 7 and used to calculate the statistics in Table 8. The bolometric method undoubtedly gives a better assessment on the evolutionary status of an embedded protostar, and we discuss the individual values for potential Class 0 protostars in the Appendix. The number of sources with high-quality submillimeter data, however, is smaller than our total infrared sample size. Since this could bias our source classifications, we retain the α_IR classifications as our main scheme. This is also done for consistency with earlier studies.

The bolometric luminosities, L_bol, of the YSOs are estimated by integrating under all available data points in each YSO's SED using a simple trapezoidal method. The distance to each YSO was taken as the distance toward its respective associated dark cloud, as discussed in Section 2. The resulting luminosities are listed in Column 7 of Table 7. Yonekura et al. (1997) estimated the IRAS luminosity function of YSO sources associated with the molecular cores in their CO survey of Cepheus. They parameterized the number of sources dN in the luminosity interval L_* + dL_* as

$$\begin{equation} \frac{dN}{dL_{*}} = N_{0} \left(\frac{L_{*}}{L_{\odot }} \right)^{-p}, \end{equation} \tag{ 2 }$$

where N₀ is a normalization factor and p is the power-law index of the function. Yonekura et al. (1997) fit N₀ = 8.9 and p = 1.40 ± 0.32 above their completeness limit of 1 L_☉ for their close group of IRAS sources. Figure 14 shows in its upper panel the function for the 133 Spitzer-identified YSOs. The slopes were fitted by a least-squares fit above the estimated break of log(L_*/L_☉) = −1.5. The slope of the luminosity function fitted to all sources is p = 1.58 ± 0.10. This value agrees with the upper range of the Yonekura et al. (1997) error bars. The individual regions show a range of luminosities plus a single source that is approximately 1 dex more luminous than the rest of the YSO population.

Zoom In Zoom Out Reset image size

The lower panel of Figure 14 shows the same histogram data as in the upper panel but plotted with dN rather than log(dN/d(L/L_☉)) on the abscissa. The luminosity distribution peaks at the point where the upper plot diverges from the single power law, i.e., at a luminosity of 0.06 L_☉. This Spitzer completeness limit is similar to that found for other c2d and Gould Belt regions (e.g., Harvey et al. 2007) and is 1.5 dex lower than the IRAS completeness limits (Yonekura et al. 1997).

A lot of sources will have a differing photometric coverage depending on the DSS, SCUBA, and IRAS detections. Therefore, we also calculate the more homogeneous L_IR, the luminosity integrated between 1–30 μm for all sources with a 3.6 μm detection. The distribution of L_IR is shown in Figure 14 as the dashed histogram, and the individual values are listed in Column 8 of Table 7. There is good agreement in the peak position and width of the two histograms, but it is noticeable that the L_IR lacks L_bol's higher luminosity tail. When compared with the equivalent L_IR distribution for Serpens and Lupus, the Cepheus L_IR distribution peaks in the same place, but the breadth of the peak is wider. All show the same peak just before 0.1 L_☉ followed by a sharp drop to zero around 0.01 L_☉ (Harvey et al. 2007; Merín et al. 2008).

We conduct basic SED modeling in order to estimate the degree of circumstellar material around the Cepheus YSOs. The SED of a YSO that has been dereddened to remove attenuation from line-of-sight extinction will include contributions from several physical components. The dominant component at short wavelengths is a stellar component from the YSO itself. A second component from a lower temperature dusty circumstellar disk and envelope will appear as an excess to the stellar component at infrared and millimeter wavelengths. This infrared excess can be estimated if we first estimate the magnitude of the stellar component and subtract it from the dereddened SED. In order to do this, we need to know the amount of interstellar extinction between the YSO and the observer, and we need to know the underlying spectral type of the YSO.

The spectral type of the YSO needs to be known so that a model can be used to estimate the YSO's stellar SED component. In general this information is not available for the Spitzer-discovered YSOs since this requires a targeted campaign of spectroscopic observations, not available for Cepheus. We follow the approach taken by Harvey et al. (2007, 2008), who made the simplifying assumption that the underlying spectral type of most YSOs is a low-mass K7 star, or an intermediate-mass A0 star in the case of the more luminous objects. This produced results statistically compatible with those obtained in Serpens, after an optical spectroscopic campaign had taken place (Oliveira et al. 2009).

For the Class II and Class III sources, the effect of reddening by line-of-sight extinction was removed from the SEDs by using the visual extinction toward each source calculated from its 2MASS J − K_s color under the assumptions of the aforementioned K7 underlying spectral type and a R = 5.5 interstellar extinction law (Weingartner & Draine 2001). The dereddened data are shown in Figures 12 and 13 as solid markers. A NEXTGEN K7 profile (Hauschildt et al. 1999) normalized against the dereddened J Band flux is plotted for comparison (the solid gray line). For a subset of sources, the dereddened optical photometry was significantly higher than the K7 profile that had been normalized to the dereddened J band flux. These sources include the known variable stars FT Cep, FU Cep, EH Cep, and FV Cep. For these sources, it was necessary to modulate the estimated A_V down by a factor of 2–4 to make the K7 profile and the dereddened fluxes coincident.

The dashed gray line in Figures 12 and 13 is the median SED for a T Tauri star in Taurus (Hartmann et al. 2005) that has been normalized to the dereddened 2MASS J flux. The profile represents a prototypical optically thick accretion disk surrounding a T Tauri star. The SEDs of some YSO candidates (e.g., YSO #3) follow this profile quite closely, but others are closer to the K7 profile and only show a small amount of infrared excess at 24 μm (e.g., YSO #22). Approximately 30 YSOs (25% of the disk population) have infrared excesses well below the median SED of the T Tauri stars in Taurus, implying that these sources have evolved or settled disks. This relatively low fraction also implies that the remaining disks in Cepheus show characteristics of being actively accreting and optically thick.

The amount of infrared excess can now be estimated for each YSO by subtracting the normalized K7 template (the solid gray line in Figures 12 and 13) from the dereddened SED data points (the filled circles in Figures 12 and 13). The luminosity of the YSO, L_star, was estimated by integrating the normalized stellar template after it had been interpolated to the observed wavelengths. The luminosity of the circumstellar disk, L_disk, was estimated by integrating the infrared excess at each of the observed wavelengths. Tables 13 and 14 list the SED modeling parameters and results for the Class II and Class III YSO candidates. Column 1 lists the YSO index. Column 2 lists the A_V that was used to deredden the SED and Column 3 lists the waveband that was used to normalize the stellar template. Columns 4 and 5 list L_star and log(L_disk/L_star).