MIPSGAL: A Survey of the Inner Galactic Plane at 24 and 70 μm

The first Spitzer survey of the Galactic plane is the Galactic Plane Infrared Mapping Survey Extraordinaire (GLIMPSE; E. Churchwell PI; Benjamin et al. 2003). GLIMPSE was one of the six original Spitzer Legacy class programs and covered two regions of the Galactic plane spanning Galactic latitudes of |b| < 1° and Galactic longitudes of 65° < l < 10° and 295° < l < 350°. GLIMPSE mapped the fields with the Infrared Array Camera (IRAC; Fazio et al. 2004a), which covers the wavelength range of 3–9 μm in four spectral bands. The original survey has been followed up by two more programs (GLIMPSE II and GLIMPSE 3D). The GLIMPSE II (PID 20201; E. Churchwell PI) survey continued the original survey for Galactic longitudes |l| < 10°, while GLIMPSE 3D (PID 30570; R. Benjamin PI) extended the latitude coverage at specified Galactic longitudes. The Galactic center region is covered by a separate IRAC program (PID 3677; S. Stolovy PI; Stolovy et al. 2006). Enhanced GLIMPSE data products and documentation are available through the Infrared Science Archive (IRSA).¹⁴ Related large Galactic plane-mapping programs that do not overlap with MIPSGAL are the IRAC survey of Vela-Carina (PID 40791; S. Majewski PI), and the MIPS and IRAC surveys of Cygnus X (PID 40184; J. Hora PI) and 21 deg² of the outer Galaxy (102° < l < 109°, 0° < b < 3°; PID 50398; S. Carey PI).

The structured and varying background and high density of point sources in the Galactic plane result in sensitivity or confidence limits for extracted point sources that are a function of position and local background. As a result, a straightforward application of σ threshold is not an adequate description of the confidence level of a source. Robitaille et al. (2007) present limits of 0.5, 0.5, 2.0, and 5.0 mJy for confident measurements of point sources at 3.6, 4.5, 5.8, and 8.0 μm, respectively, for the GLIMPSE data set. The Robitaille limits are determined from examination of several thousand spectral energy distributions and corresponding residual images. The GLIMPSE point-source catalog (Meade et al. 2007) is even more restrictive and used simulated truth sets and analysis of a deeper verification field to determine their sensitivity limits for the very high-reliability catalog (99.99%). For MIPSGAL, we currently estimate that sources brighter than 2.5 mJy at 24 μm are most certainly real. This value was determined by inspection of source-fit residuals, matches to existing truth catalogs, and reliability studies using injection of artificial sources to the MIPSGAL mosaics (S. Shenoy et al. 2009, in preparation). The GLIMPSE data set is more sensitive to stellar photospheres than the MIPSGAL survey. The final GLIMPSE catalog reaches a limiting magnitude of 14.1 mag at 3.6 μm (Meade et al. 2007) compared to our most optimistic cutoff of 9 mag at 24 μm for stars without infrared excess (bare photospheres). However, for sources redder than the main sequence, MIPSGAL is generally more sensitive than GLIMPSE. As a result, a greater fraction of sources detected in MIPSGAL will be evolved stars and protostars. For many lines of sight, GLIMPSE is point-source confusion limited due to the high stellar-column density in the Galactic plane, while MIPSGAL is generally limited by the structure of the diffuse background. In addition, the Galactic midplane has a significant opacity even at mid-infrared wavelengths. Mercer et al. (2005) showed the path length that the GLIMPSE survey was able to probe decreased with decreasing Galactic longitude. As the extinction at 24 μm is 40% lower than that at 8.0 μm (Draine 2003), MIPSGAL observations for highly extincted lines of sight will provide complementary information on Galactic structure even though the 24 μm data are less sensitive to the red-clump giant population used as a probe of Galactic structure (Benjamin et al. 2005). As discussed in § 2.1, the combination of IRAC and MIPS photometry will be a powerful tool for identifying and characterizing protostars.

The AKARI mission has recently completed an all-sky survey in the mid-infrared at 9 and 18 μm and in the far-infrared from 60 to 180 μm. The sensitivity of their 18 μm band is 120 mJy per survey scan, and it has a resolution of 10'' for the all-sky survey observations (Onaka et al. 2007). The Wide-Field Infrared Survey Explorer (WISE; Mainzer et al. 2005) is currently scheduled to launch in late 2009 and will survey the sky in four infrared bands, 3.3, 4.7, 12, and 23 μm with a resolution of 12'' and nominal sensitivity of 2.6 mJy at 23 μm. Due to the survey strategies of both AKARI and WISE, their final sensitivities are a function of ecliptic latitude. The region of the Galactic plane observed by MIPSGAL spans ecliptic latitudes from -52° to +40° with the survey longitude edges being at the ecliptic latitudes farthest from the ecliptic plane. The AKARI and WISE sensitivities increase by a factor of ∼1.2 at the MIPSGAL survey edges. The AKARI and WISE observations complement the higher-resolution and higher-sensitivity observations at 24 μm from MIPSGAL. Both the AKARI and WISE teams plan to deliver high-fidelity source catalogs to the public and the WISE team plans to distribute calibrated mosaics.

The upcoming survey of the northern Galactic plane with SCUBA-2 on the James Clerk Maxwell Telescope (JLS/GPS; Holland et al. 2006) and the recently approved survey of the inner Galactic plane with Herschel (Hi-Gal; Molinari et al. 2005) will provide important, long wavelength complements to the existing mid-infrared surveys of the Galactic plane. The JLS/GPS survey will cover most of the first quadrant region observed by MIPSGAL/GLIMPSE with comparable resolution at 450 μm to the MIPSGAL 24 μm data. Hi-Gal will cover the entire MIPSGAL/GLIMPSE survey region at 75, 175, 250, 350, and 500 μm with equivalent resolution at 75 μm to MIPSGAL at 24 μm. The APEX Telescope Large Area Survey: the Galaxy (ATLASGAL) is an 870 μm survey using the LaBoCa camera on the Atacama Pathfinder Experiment (APEX) of the plane between -80° < l < 80°, |b| < 1° with 18'' resolution. At this time, 95 deg² of the ATLASGAL fields have already been mapped (Menten et al. 2008). A continuum survey at 1.3 mm (31'' resolution) of the northern Galactic plane, which has considerable overlap with GLIMPSE/MIPSGAL, is currently being conducted with Bolocam on the Caltech Submillimeter Observatory (Williams et al. 2007). Other submillimeter surveys that have some overlap with the MIPSGAL survey field are the Balloon-Borne Large Aperture Submillimeter Telescope (BLAST; Pascale et al. 2008), which surveyed several fields of a few square degrees in size at 250 (32'' resolution), 350, and 500 μm overlapping the MIPSGAL maps (Chapin et al. 2008), and the Archeops balloon mission (Benoît et al. 2002), which covered the Galactic plane between 30° < l < 180° (Désert et al. 2008) at wavelengths of 550 (480'' resolution), 875, 1380, and 2100 μm. These longer-wavelength surveys will provide critical photometric data on the earliest stages of star formation and identify the mass reservoirs out of which the infrared-identified massive protostars and protoclusters are forming. For studies of the distribution and energetics of interstellar dust, the far-infrared and submillimeter provide information on large dust grains (BGs) in a complementary fashion to the 8 μm data from GLIMPSE that are sensitive to the emission from polycyclic aromatic hydrocarbons (PAHs) and the 24 μm data of MIPSGAL that trace the diffuse emission from very small grains (VSGs).

The 24 and 70 μm images from MIPSGAL are powerful tracers of class 0–III protostars. Class I–III protostars have spectral energy distributions (SEDs) that peak between these wavelengths. The modeling of Whitney et al. (2003) demonstrates that protostars are easily distinguished from other objects in color-color plots using 70 and especially 24 μm as one of the bandpasses. In combination with the mid-infrared photometry from IRAC and near-IR photometry of 2MASS, sources can be compared to sophisticated model spectra (Robitaille et al. 2006) to probe evolutionary state and mass. Using the models of Whitney et al. (2003) as a guide, an isolated 1 M_⊙ protostar can be detected in the MIPSGAL-GLIMPSE data out to the distances given in Table 1. However, many low-mass stars will not form in isolation and many of the sources identified will be blends and portions of protoclusters. While distributed low-mass protostars (e.g., Allen et al. 2008) can be individually detected and photometered, only the aggregated infrared emission will be available for many sources.

The sensitivity and resolution of MIPSGAL at 24 μm permit resolution of ultracompact H II (UCHII) regions to a distance of ∼4 kpc assuming a typical UCHII diameter ∼0.1 pc (Wood & Churchwell 1989) and detection of UCHIIs throughout the Galactic disk. The slight mid-IR excess from very early, massive (M > 5 M_⊙) stars will also be detectable as even cores of M = 10⁴ M_⊙ and T = 25 K (pre-protostellar cores) produce enough 24 μm emission to be detected at a distance of 3 kpc (Sridharan et al. 2002). The 70 μm data are sensitive to even younger objects albeit at a lower spatial resolution.

These data can be used to examine the clustering of protostars as a function of mass and evolutionary state, extending the sample of well-characterized star-forming regions beyond 1 kpc. It is possible that if clustering is studied for a statistically significant sample of regions, current (competing) theories of massive star formation such as turbulent fragmentation (Krumholz et al. 2005) and competitive accretion (Bonnell et al. 2004) can be tested. Shocks and winds from nearby massive stars have been shown to trigger star formation (e.g., Redman et al. 2003), but the efficiency of triggers relative to other factors has not been quantified. By surveying most of the star formation in the Galactic disk, MIPSGAL, GLIMPSE, and the ongoing and future surveys mentioned in §§ 1.2–1.3 have the potential to identify the key processes in the formation of massive stars.

The combination of these infrared surveys with spectroscopic surveys of molecules (e.g., ¹³CO from the Galactic Ring Survey; Jackson et al. 2006) and radio recombination lines (e.g., Caswell & Haynes 1987; Lockman 1989) can provide kinematic distances permitting a census of massive star formation in the inner Galaxy. By counting the number of high-mass protostars, a direct measure of the current star-formation rate can be made. This measurement can then be compared to less direct measurements such as total far-infrared luminosity, which is currently used to estimate the star-formation efficiency of external galaxies. The synthesis of surveys over the Galactic plane at varied wavelengths can be used to calibrate extragalactic star-formation efficiency and activity tracers.

Mid- to far-infrared observations directly measure the emission of several different dust components. Observations shortward of 24 μm are often dominated by emission from PAHs or other carriers of the infrared emission bands. The 24 μm passband is sensitive to emission from very small, stochastically heated dust grains while longer wavelengths provide information on BGs which are closer to thermal equilibrium (see Draine 2003 for a review of interstellar grains).

The SED of dust emission is a sensitive function of the relative populations of these three grain types and the radiation field they are subjected to (Flagey 2007). In combination with the GLIMPSE data, our survey at 24 and 70 μm can be used to infer the column density and the excitation conditions of PAHs, VSGs, and BGs along a line of sight. The addition of longer-wavelength data permits more detailed studies of the large-grain population. From broadband color variations, changes in ambient radiation field and ionization state can be observed. In turn, variations in the incident radiation field can be used to infer the nature of the exciting sources of the dust.

For fixed dust component abundances, we can model the dust emission for different incident radiation field intensities as shown in Figure 2. We use the spectral shape of the standard interstellar medium radiation field (ISRF) from Mathis et al. (1983) and scale it by a factor (χ) of 1, 100, and 10000. As the energy density of the ISRF increases, the BGs' temperature increases and their emission shifts to shorter wavelengths. The peak emission of the VSG component shifts less significantly to shorter wavelengths while the PAH emission does not shift at all. As a consequence, the color ratios between IRAC 8 μm, MIPS 24 μm, and MIPS 70 μm significantly depend on the ISRF. In the diffuse interstellar medium (ISM; χ ∼ 1), the BG temperature is about 15–20 K and the 24 μm to 70 μm ratio is smaller than 1. As long as the radiation field is not strong enough to make the BGs' contribution within the 24 μm passband significant, the 24 to 70 μm ratio decreases. For highly excited regions (χ ∼ 10⁴), the BGs can be hotter than 100 K and the SED reaches its maximum at wavelengths close to 24 μm. The 24 to 70 μm brightness ratio thus increases and can be much higher than 1 (e.g., in supernovae remnants). For a given or known incident radiation field, IRAC 8 μm, MIPS 24 μm, and MIPS 70 μm relative intensities can be used to directly probe the relative abundances of each dust component.

Zoom In Zoom Out Reset image size

The 24 μm data can be used to map infrared extinction using the diffuse background in the Galactic plane. Dense molecular cloud cores can have extinctions of 1–2 mag at 24 μm (Carey et al. 1998). Several thousand IRDCs (mid-IR extinctions >1 mag) have been identified in the MSX and ISOGAL survey fields (Perault et al. 1996; Egan et al. 1998; Simon et al. 2006). The 8/24 μm opacity ratio of these objects can be used to infer the grain size distribution (e.g., Hennebelle et al. 2001) and the mid-infrared optical depth of the clouds can be used to estimate the mass of the molecular cores (Carey et al. 1998) independently of spectral-line gas tracers or submillimeter continuum and without regard for depletion and/or temperature and dust emissivity effects.

The sensitivity and angular resolution of Spitzer make possible the study of ISM physics on scales ranging from the large-scale structure toward the Galactic center to those at which supernova explosions and the winds of massive stars dominate the structure down to that at which turbulence does. Maps of the diffuse 24 and 70 μm emission in the Galactic plane will constrain the physical conditions of the gas clouds in which the VSGs and BGs reside.

Where corresponding CO, H I, and radio continuum data are available, a multicomponent model (Sodroski et al. 1997; Paladini et al. 2007) can be used to partition the small-grain emission originating in each of the different ISM phases (warm ionized, warm neutral, and cold neutral media: WIM, WNM, and CNM, respectively). Due to the low resolution of earlier infrared data sets, inversion techniques (Bloemen et al. 1986) were only able to estimate dust properties by averaging over large (few kpc) Galactocentric radii bins. MIPSGAL data, combined with recent H I (Stil et al. 2006) and ¹³CO (Jackson et al. 2006) high-resolution data cubes, will make possible the derivation of emissivity coefficients, temperatures, and gas-to-dust ratios on sub-kpc scales, thereby providing a detailed picture of the spatial variations of dust properties. The inversion technique will be limited by the resolution of the H I data; currently 60'' for various components of the international Galactic plane survey (IGPS; Stil et al. 2006; McClure-Griffiths et al. 2005).

A two-dimensional power spectrum analysis of the Milky Way midplane will allow us to address new questions relating to ISM structure. For lines of sight dominated by a single emitting region, the power spectrum is a powerful tool, but interpretation of more complicated lines of sight is very challenging. Using data from the Spitzer Galactic First Look Survey, Ingalls et al. (2004) found that the power spectrum of 24 and 70 μm emission toward (l,b) = (254.5°,-9.5°) has a spectral index of -3.5 at small scales (6''–60''), consistent with Kolmogorov turbulence. At large scales, however, there is a break in the spectrum, suggesting that the emitting medium for that region is formed into sheets of width 0.3 pc. This change in spectral index would have been difficult or impossible to measure with previous far-infrared or radio surveys of the Galactic plane, which had resolutions of at best 45''. A hint of the spectral index break is shown in the data of Gautier et al. (1992); however, without the small spatial scale data, they were able to fit the power spectrum with a single power law of index -3.0. Because the cold neutral medium seen in [H I] self-absorption is deduced to be sheetlike (Heiles & Troland 2003), it is possible that the 24 μm emission mainly traces CNM gas. Another possibility, that the 24 μm emission traces only a thin surface layer of the gas, is unlikely because the break is seen also at 70 μm. With MIPSGAL, the prevalence of Kolmogorov-like turbulence in Galactic gas can be tested on scales ranging from kiloparsecs down to tenths of parsecs. In addition, the mass fraction of sheetlike versus volume-filling emitting gas can be measured.

As low- to moderate-mass stars evolve up the asymptotic giant branch, burning hydrogen and helium in shells around their cores, their atmospheres expand and cool. The stars experience significant mass loss via strong stellar winds, resulting in the expulsion of a substantial fraction of their initial mass. In the outer layers of the atmosphere and in the circumstellar envelope surrounding the star, the environment is ripe for dust formation. Indeed, AGB stars are thought to produce 50% or more of the dust in the Galaxy (recent reviews of AGB evolution, the mass loss process, and dust formation include Willson 2000, Herwig 2005, and Zijlstra 2006). As around the young stellar objects (YSOs) and protostars discussed in § 2.1, the dust around AGB stars emits primarily in the infrared. Because it is newly formed, and relatively unprocessed by UV radiation, the dust observed around evolved stars has spectral features, and hence photometric properties, distinct from those seen in young objects or in older objects such as planetary nebulae. Thus, when the MIPSGAL 24 μm data are compared with data at shorter wavelengths such as GLIMPSE and 2MASS, the evolved stars occupy a distinct region in color-color space, as recently demonstrated for the Large and Small Magellanic Clouds by Blum et al. (2006) and Bolatto et al. (2007), respectively. While the reddening toward our Galactic sources will not be as uniform as in those galaxies, the old and young stellar populations should still be distinguishable. Robitaille et al. (2008) have used MIPSGAL 24 μm data in conjunction with the GLIMPSE point-source catalog to identify ∼7000 candidate AGB stars.

Unlike IRAC images, no pointing refinement is performed on the MIPS BCDs. The pointing accuracy of the mosaics is the result of the combined effect of the blind pointing accuracy of Spitzer (≲1^'' rms) and the inability of the MIPS Cryogenic Scan Mirror Mechanism to return to the same zero point (∼0.5^''). An assessment of the astrometric accuracy is given by matching positions of point sources in the 24 μm mosaics with the sources at 8 μm from the GLIMPSE Point Source Catalog.²² Figure 9 displays the offset between source matches at 24 and 8 μm. The median offset between 24 μm sources and their 8 μm counterparts is 0.85'', which is in keeping with the two effects mentioned previously. This offset is small compared to the FWHM of MIPS point sources but is more significant compared to the ∼2^'' FWHM for point sources at 8 μm imaged by IRAC. GLIMPSE sources were matched by finding the nearest GLIMPSE source within a 5'' search radius. This search radius balances the need to have a large enough radius to identify any systematic pointing error in the MIPSGAL data set with reducing the number of spurious source matches. A straight positional association was used without a flux criterion because sources with widely varying spectral energy distributions are detected at both 24 and 8 μm. The estimated random matches as a function of distance between sources assuming an average source density of 33,225 sources per square degree for GLIMPSE is also plotted using a dashed line. The tail of the distribution between 1.5'' and 3'' is consistent with outliers from the combination of the errors in the initial acquisition slew and scan mirror zero point. We anticipate correcting the misregistration of the MIPSGAL data in a future version of the data products.

Zoom In Zoom Out Reset image size

We checked the photometric accuracy of our data reduction by comparing the surface brightness of the data to MSX 21.3 μm measurements. Figure 10 displays the pixel-to-pixel correlation between MIPSGAL data smoothed to the same resolution as MSX 21.3 μm data of M16. Each data set was color-corrected using a correction appropriate for the ISM in a massive star-forming region such as M16 (Flagey 2007). The pixel-to-pixel relation is well matched by the relationship

where S_MIPSGAL and S_MSX are the measured surface brightnesses, 0.99 and 1.03 are the color corrections, and λ_MIPSGAL and λ_MSX are the isophotal wavelengths of MIPS 24 μm and MSX Band E, respectively. The relationship is appropriate for the spectral energy distribution of dust in a massive star-forming region. The departure of the data from the model for MIPSGAL surface brightnesses starting at 1000 MJy sr^-1 and more significantly for 1500 MJy sr^-1 is most likely due to bright compact sources that will have a different spectral energy distribution than the ISM. At 1700 MJy sr^-1, the plot clearly indicates that the MIPSGAL data are saturated. The turnover in the slope is due to regions that are saturated in MIPSGAL being compared to unsaturated MSX regions. The saturated MIPSGAL pixels are set to not-a-number (NaN), and when summed over the MSX resolution element (20'') the saturated pixels are ignored—resulting in a lower surface brightness for the MIPSGAL data.

Zoom In Zoom Out Reset image size

As a quick check of the point-source photometry that can be derived from the 24 μm mosaics, we measured the flux densities at 24 μm for 27 stars of known spectral types (K0III–K4III and A0V–A1V) and with flux densities from 2MASS and GLIMPSE. The stars were chosen to be in moderate to low structure background regions. The set is not comprehensive in nature and a more detailed analysis is in S. Shenoy et al. (2009, in preparation). The APEX source extractor was used on the mosaics in a manner similar to our source-extraction pipeline. The fit residuals were inspected by eye and were good for all of the sources despite a fair amount of background structure in the images. Figure 11 displays the 24 μm image of one of the stars used (HD 166290) and the residuals of the source fit.

Zoom In Zoom Out Reset image size

We predicted the MIPSGAL flux densities for the stars using the GLIMPSE 8.0 μm measurement and a spectral template of a well-characterized star of the same spectral type. We measured the color differential (C[8 - 24]) between the predicted [8] - [24] color from the spectral standards and the measured [8] - [24] color from the MIPSGAL and GLIMPSE photometry. That is,

where ([8] - [24])_s is the [8] - [24] color of the standard with the same spectral type and ([8] - [24])_d is the [8] - [24] color measured from the data.

The spectral templates used were from Engelke et al. (2006) for the K0–K2 giants (except for K1III) and Cohen et al. (2003) for the A dwarfs, K1, K3, and K4 giants. Table 3 displays the template magnitudes for the spectral types used. In converting the flux densities to magnitudes, we used zero-magnitude flux densities of 64.13 Jy (Reach et al. 2005) and 7.17 Jy (Engelbracht et al. 2007) for [8.0] and [24.0], respectively. Figure 12 plots the color differential (C([8] - [24])) of the measured photometry compared to the model prediction for the 27 stars with spectral types corresponding to the available spectral templates and good photometry in MIPSGAL and GLIMPSE 8 μm. Most of the stars have photometry in good agreement with the models, although four stars have significant excesses (>0.2 mag) at 24 μm. Three of the excess stars are A dwarfs and most likely have debris disks. The fourth star, HD 168701, is an eclipsing variable with a secondary of unknown spectral type (Malkov et al. 2006). The average excess for the stars with C([8] - [24]) > -0.2 is +0.03. A comparable offset has also been noted by the Surveying the Agents of Galaxy Evolution (SAGE) Legacy team using a similar analysis.²³ If the revised 24 μm calibration factor of Engelbracht et al. (2007) is used, then the excess is reduced to +0.01. In any event, the offset is within the cross-calibration error (5%) between IRAC and MIPS.

Zoom In Zoom Out Reset image size

A portion of the data for |β| < 15° has been inspected for asteroids by differencing the multiple epoch data and looking for sources that are paired with a negative source in the difference mosaic. Figure 13 displays the difference image of two overlapping epochs of data. Several asteroids are visible in the difference image. The slight pointing offset between scan legs is also present as a positive-negative dipole residual for stationary point sources in the difference mosaic. Most of the offset is in the in-scan direction, while asteroid motion is in the cross-scan direction (which is roughly aligned with ecliptic longitude). The magnitude of the in-scan offset is typically a fraction of an arcsecond.

Zoom In Zoom Out Reset image size

At the present time, we have examined in detail 33.3 deg² of the multiple-epoch–covered regions. There are 1189 single-epoch transient detections corresponding to 601 sources in the areas analyzed. Of these sources, 213 are objects in the Horizons²⁴ solar system database. Unfortunately, 75% of the single-epoch objects remain in both the final mosaics and the current point-source catalog. A list of these sources and their positions will be made available as soon as the analysis is complete.

For β = 0°, ∼40 asteroids are detected per square degree. At the ecliptic latitude limit of our multiple-epoch data, β = 15°, there are ∼10 asteroids per square degree. If we make the overly conservative assumption that the asteroid number counts do not fall off with ecliptic latitude beyond β = 15°, then for the remaining 200 deg² there should be at most 2000 unidentified asteroids in the mosaics for l > 15° and l < 350°. In the inner 16 deg², the observations of PID 20414 did not employ a multiple-epoch observing strategy; therefore, asteroids have not been rejected in most of the Galactic center field and of the order of 600 asteroids probably remain in the mosaics of that field. Due to the overlap at the edges of AORs in PID 20414, some asteroid rejection has occurred in the Galactic center field. Users of the image and catalog products should be careful in interpreting sources that are detected only at 24 μm.

After preprocessing the BCDs (Mizuno et al. 2008), the MOPEX package (Makovoz et al. 2006) is used to generate the image plates. The mosaicking process resamples the BCDs onto a common sampling and projection. Outlier rejection is performed to remove radiation hits and other transients. The outlier rejection uses the median absolute deviation (MAD) to estimate the variance of the input data going into each output pixel. Data samples more than +5 or −3 σ above or below the median were omitted as outliers. The rejection was positively skewed to prevent flux from being removed in the wings of point sources. In the presence of Poisson noise, MAD will tend toward the lower pixel values for a distribution that is bimodal. The outlier rejection also removes some of the asteroids from the mosaics with multiple-epoch observations. For the brighter asteroids, the wings of the PSF below the thresholding are not rejected and have to be masked by hand. Currently, this masking is incomplete in the mosaic products.

The 24 μm mosaics are 1.1° × 1.1° in size and are spaced on 1° centers. The images are in units of surface brightness (MJy sr^-1). A model of the Zodiacal light (Kelsall et al. 1998) has been subtracted from the input data before co-addition. Pixels without data contain IEEE floating point NaNs. Approximately 0.09% of the survey region is saturated and 0.03% is blanked out due the editing of bright source artifacts. The images are sampled on 1.25'' pixel centers. The image coordinates are Equatorial J2000 and a tangent projection is used in constructing the mosaic. A rotation angle is applied (nonzero CROTA2 keyword) such that the images are aligned in Galactic coordinates. This is strictly true for the center of each mosaic and holds true for all practical purposes for the entirety of each plate. The astrometric information is encoded in each mosaic header following the format recommendations of Calabretta & Greisen (2002).

The mosaic headers include information on the processing steps applied as well as a list of the AOR request keys contributing data to the mosaic. The number of AORs used in the mosaic is given by the NAOR keyword. The AOR request keys are provided by the AORKEY## keywords (where ## is an integer, 00, 01, 02, ...). The start time of each AOR used is contained in the AORTIM## keywords. The processing version of the mosaics in this delivery is noted in the keyword VERSION and is 1.1 for the l > 10° data and 1.2 for the l < 10° and fourth quadrant data. An example of a mosaic header is given in the Appendix.

In addition to the astrometric and photometric keywords, including the flux conversion factor (FLUXCONV) and gain (GAIN), the mosaic headers include a set of Boolean flags to indicate whether various artifact-mitigation steps were performed on the input BCDs. Corrections for array droop, jailbars, dark latents, and bright latents (see Mizuno et al. 2008 for more details) are indicated by the DROOP, JAILBAR, DARKLAT, and BRITELAT keywords. Removal of a bias pattern in the BCDs is indicated by the WASHBRD keyword (as the bias pattern has a washboard appearance in the BCDs). Overlap correction is indicated by the OVERLAP keyword. Flagging of glints due to bright sources on the edge of the array and bright, known asteroids is indicated by the EDGEMASK and ASTRDMSK keywords.

The image header also includes various statistics such as the minimum and maximum data values, keywords DATAMIN and DATAMAX, respectively. The input pixel coverage statistics are included as are those for the coverage mosaics (see § 5.5). The number of pixels in the input BCDs that are hard saturated (BCD pixels that are saturated before the second sample in the data ramp) and percentage of mosaic pixels that are flagged as hard saturated are given in the keywords NHARDALL and PHARDSAT. The keywords NSOFTSAT and PSOFTSAT provide the analogous quantities for soft saturation (pixels that are saturated by the third sample in the data ramp).

As part of the mosaic process, uncertainty mosaics in units of surface brightness (MJy sr^-1) are also generated. The uncertainty mosaic headers include the same astrometric information as the mosaics. The value of each pixel in the uncertainty mosaics is the standard deviation of the interpolated BCD pixels that are co-added into the output mosaic pixel divided by the square root of the coverage of the input BCD pixels. In general, the standard deviation is close to the weighted average of the BCD uncertainties for extended structures and background regions. Both the BCD uncertainties and the standard deviation of the pixel stacks are in accord with the Poisson-dominated noise as expected for the high backgrounds observed. For point sources, the standard deviation is several times higher than the noise estimate from the BCD uncertainties. The larger variance is due to the small astrometric misalignment between BCDs which blurs point sources slightly and contributes to the scatter in the values of the pixel stack.

The coverage mosaics indicate the number of input image pixels co-added for each mosaic pixel. The mosaicking procedure weights each input pixel by its area overlap with an output pixel; therefore, the coverage can include fractions of a pixel. On average, 12 input data samples contribute to each output mosaic pixel. The coverage is highly position dependent and varies from a nominal minimum of 9–10 up to 18–19 for the regions where edges of scans overlap. The coverage mosaics also include the same astrometric information as the image mosaics. The keywords COVRGMAX, COVGRMIN, and COVRGMN, give the minimum, maximum, and mean coverage of all pixels with nonzero coverage, respectively. NCOVREG0 provides the number of pixels with zero coverage.

The mask cubes display the number of input samples that were flagged as problematic for each output mosaic pixel. None of the flagged samples were used to create the mosaics. Positions that have no good data are filled with NaN in the output mosaic. Therefore, the mask cubes are complementary to the coverage mosaics and indicate why the coverage for a given pixel is less than optimal. Seven states are indicated in the mosaic mask. The first plane indicates input pixels that are hard saturated. The second plane indicates BCD pixels that are affected by stray light from a bright source on the edge of or just off the array. The third plane flags input pixels that cannot be flat-fielded. The fourth plane displays pixels that have been edited out due to known asteroids or asteroids that were partially removed by the outlier rejection step for positions observed at multiple epochs. The fifth plane indicates data that are flagged as bad in the MIPS pmask (see the MIPS Data Handbook²⁵ for more details). The sixth plane flags BCD pixels that are soft saturated. The final bit plane displays regions with missing input pixels due to downlink problems.

The mask cubes have the same astrometric information as the image mosaics. The photometry of regions with some saturated samples will probably be less reliable. Faint sources or structure in regions with known stray light should be investigated further to ensure their validity.

Figure 14 shows an IRDC, G11.11−0.11, at 24 and 70 μm. Extinction at 24 μm is commonly seen in these dense molecular cores; however, there had only been one previous measurement of extinction at longer wavelengths (Lis & Menten 1998), for the IRDC G0.26−0.01. That particular IRDC is near the Galactic center and has been considered atypical of the general population of IRDCs. We estimated the opacity for G11.11−0.11 by assuming that it is a simple extincting slab without a source function at 24 and 70 μm. The foreground emission is estimated to be 21% of the total along the line of sight, and the background is estimated by performing a planar fit of the surface brightness across the IRDC from the region around the cloud. The estimated peak opacity at 70 μm is 0.28 and is 1.1 at 24 μm. Using a gas-to-dust mass ratio of 1.05 and the R_v = 5.5 extinction law of Weingartner & Draine (2001), we derive a peak column density of 4.6 × 10²² cm^-2 from the 24 μm opacity estimate. This finding confirms the previous estimates of column densities for this IRDC (Carey et al. 2000). The ratio of 70 to 24 μm opacity is 2.5× that predicted by the Weingartner & Draine (2001) model and is probably more consistent with the extinction produced by grain growth through aggregation and/or accretion of icy mantles (Ossenkopf & Henning 1994).

Zoom In Zoom Out Reset image size

Figure 15 is a three-color image of two star-forming columns at l ∼ 60° identified in the MIPSGAL and GLIMPSE surveys. The columns are ∼30^' east of the young stellar cluster NGC 6823. These columns are similar in morphology to the "pillars of creation" seen in M16 (Hester et al. 1996) and the "mountains of creation" in W5 (Allen et al. 2005). Star-forming columns and "elephant-trunk" globules are prevalent throughout the Galactic plane as noted in previous infrared surveys of the plane such as MSX. The 8 and 24 μm data show the expected presence of young stellar objects at the tips of the columns. Also seen at 24 μm are faint point sources in the middle of each pillar. These sources are much more apparent at 70 μm, suggesting that they are even younger protostellar objects. These observations support a scenario where star formation is ongoing in the pillar and not induced by the radiation pressure and winds from the nearby hot star at the pillar/wind boundary. Indebetouw et al. (2007) reached a similar conclusion for their study of M16. This scenario is in contrast to the recent result by Ubeda & Pellerin (2007), where the mid-infrared data support a triggered star-formation scenario. Many other cases of both revealed and triggered star formation are present in the MIPSGAL data set.

Zoom In Zoom Out Reset image size

A preliminary search for infrared counterparts to supernova remnants (SNRs) listed in Green (2006) has revealed nine counterparts out of the 30 candidates surveyed so far. Out of the nine, three objects have a low confidence level of detection yielding a conservative detection rate of 20% . This value is roughly similar to the rate reported for the GLIMPSE survey of Galactic supernova remnants (Reach et al. 2006) and six of the nine SNRs with 24 μm detections are also detected at 8 μm. As an example, we present in Figure 16 the infrared detection of the supernova remnant G27.4+0.0 (Kes 73), which is also known to have a compact X-ray source (Helfand et al. 1994). This particular object is shown at GLIMPSE 8 μm, MIPSGAL 24 and 70 μm, VGPS 21 cm (Stil et al. 2006), and MAGPIS 90 cm radio continuum (Helfand et al. 2006). The lack of emission at 8 μm and the faint glow at 70 μm suggests that its 24 μm detection is mostly due to nebular emission lines such as [O IV] (25.89 μm) and [Fe II] (25.99 μm) with a small contribution to the continuum emission from dust produced in the remnant. A portion of the 70 μm emission is possibly due to [O I] (63 μm). These observations and the recently revised age of 500–1000 years for Kes 73 (Tian & Leahy 2008) suggest that little dust has formed in this remnant which is in contrast with the results of Rho et al. (2008) for the Cas A SNR.

Zoom In Zoom Out Reset image size

Small (< 1^') ring and shell morphologies are pervasive through the Galactic plane in the mid-infrared. Over 100 of these objects have been identified in casual visual inspection of the MIPSGAL mosaic images. The density of small rings or shells is about 0.5 deg^-2 in the inner Galaxy; however, their distribution is nonuniform. Some examples are shown in Figure 17. In general, these objects are smaller than the stellar wind-blown bubbles of Churchwell et al. (2006, 2007), but there is overlap between these two sets of sources.

Zoom In Zoom Out Reset image size

Many of these objects are not apparent in 3.6–8.0 μm data from the GLIMPSE survey. It is impossible to tell without an infrared spectrum whether the absence of 8 μm emission is due to a lack of corresponding PAH and hotter dust emission if the 24 μm objects are dusty or the presence of 24 μm emission is due to line emission from ionized gas. The GLIMPSE and MIPSGAL sensitivities are well matched for ISM emission, so it is unlikely that an object with a standard ISM composition of dust grains would produce 24 μm emission without measurable emission at 8 μm.

The likely suspects for these objects are some type of evolved star; whether a low- or intermediate-mass AGB or a massive luminous blue variable (LBV), distant Wolf-Rayet star, or possibly a planetary nebulae. In MSX data, four candidate LBV stars had well-defined diffuse rings surrounding a point source (Egan et al. 2002; Clark et al. 2003) that are very similar in morphology to the MIPSGAL rings. The ring or bubble of emission in these cases would be from hot, small dust grains in the stellar winds and ejecta. Potentially, these blobs could comprise the "missing" massive evolved stars believed to be present in the Galaxy (Shara et al. 1999).

However, not all shells follow this paradigm. Morris et al. (2006) determined that a particular diffuse 24 μm-only blob consisted mainly of [O IV] line emission without continuum emission from dust. Their interpretation was that the source was most likely a distant SNR. If a significant fraction of the bubbles are similar to the Morris object, then it is quite likely there is a new and poorly understood class of supernova remnant.

We have searched for stars with infrared colors indicative of debris-disk candidates in a preliminary version of the MIPSGAL 24 μm point-source catalog for the longitude range 10° < l < 60°. Debris-disk candidates were identified by band-merging the MIPSGAL point-source catalog with the 2MASS and GLIMPSE catalogs. A simple positional association using a 3'' search radius was used. Sources with multiple associations between catalogs were dropped. We have used IRAC color-magnitude and color-color criteria to reject extragalactic contamination (Fazio et al. 2004b) including PAH-rich galaxies (Hernández et al. 2007). Only sources with 2MASS and IRAC colors representative of stellar photospheres, that is,

were considered. These color criteria select 16.3% (58,804 out of 360,038) of the band-merged point sources. 8454 sources with 0.6 < K - [24] < 3.3 are identified as debris-disk candidates. Future work will determine the distribution of these sources as a function of spectral type and age.