THE SPITZER–WISE SURVEY OF THE ECLIPTIC POLES

The EP-N data are comprised of two separate AORs: Key = 23820800 (2007 August 16) and 23821056 (2007 September 13), whose separation in time was chosen to optimize the spacecraft rotation between epochs. The region was mapped with 12s high dynamic range imaging using an 18 × 18 exposure grid with one-half frame overlaps, covering a total area of approximately 0.4–0.5 deg² with a coverage depth of eight individual exposures (four from each epoch). The 8 × redundancy scheme improves sensitivity, artifact removal and mitigation of focal-plane non-uniformities. The individual basic calibrated data (BCD) frames were combined to form deep mosaic images using the WISE Astronomical Image Co-adder (AWAIC; Masci & Fowler 2009), which includes removal of temporal outliers (e.g., cosmic rays), background matching, and astrometric and photometric calibration. The total integration time is 96s (8 × 12s), achieving 5σ point-source limits of 10, 12, 27, and 45 μJy (18.6, 17.9, 16.6, and 15.4 mag), in IRAC-1, 2, 3, and 4, respectively. The EP-N mosaic is shown in Figure 2; note the bright sources near the center of the field: NGC 6543, the "Cat's Eye" Nebula and the Seyfert Type-II galaxy NGC 6552 (Figure 3). The non-standard orientation of the EP images reflects the native orientation of the spacecraft as it mapped the poles, thereby producing the best spatial resolution mosaics.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The EP-S data are comprised of two separate AORs: Key = 24321024 (2007 October 19) and 24320512 (2008 January 5). As with the EP-N, the epoch separation in time was chosen to optimize the spacecraft rotation. A smaller region of the area was mapped, with grid size 15 × 17, also with half-image overlaps and 8× redundancy. For these dedicated and relatively small area observations the depth is comparable to the EP-N survey and roughly twice the depth of the SAGE imaging of the LMC (Meixner et al. 2006). To construct a larger area field, we combine our data with that of SAGE, creating an area of 1.28 deg², thus fully covering the WISE CVZ in the south (Figure 4). The SAGE region has 5σ limits of 20, 25, 65, and 85 μJy (17.9, 17.1, 15.6, and 14.7 mag) in IRAC-1, 2, 3, and 4, respectively; bear in mind that the EP-S coverage is not uniform throughout but 2× deeper in the northwest section compared to the southeast section that is comprised of SAGE imaging (see Figure 4).

Zoom In Zoom Out Reset image size

The EP-N data are collected in AOR Key = 24320768 (2008 January 6), and the EP-S data in AOR Key = 24320512 (2008 January 5). Both polar regions were mapped using a fast scanning mode, 12 legs of one degree length and 5' width. Only the MIPS-24 imaging is reported here, but the MIPS-70 imaging exists in the Spitzer archive. The individual BCD frames were combined to form deep mosaic images using WISE-AWAIC, with the pixel scale and orientation matched to that of the IRAC mosaics, thus oversampling the native pixel scale while maintaining the photometric calibration and spatial resolution integrity of the original data. Additionally, MIPS photometry-mode observations of four candidate calibration stars were conducted. The basic Spitzer Science Center pipeline products were used to extract photometry from these four sources. The achieved 5σ point-source sensitivity is 325 μJy (10.90 mag) for both the EP-N and EP-S.

Source detection and characterization were carried out using software tools adapted from the WISE data processing pipeline developed by the WISE Science Data Center (WSDC) of the Infrared Processing and Analysis Center (IPAC). The distinguishing characteristic of the system is that it uses all (five) IRAC+MIPS-24 bands simultaneously detect sources. The detection is based on the thresholding of a "detection" image derived from a set of matched filter images in the relevant bands. The algorithm and performance are presented in Marsh & Jarrett (2011); key algorithmic details include calculating the optimal matched filter at each wavelength and combining the resulting single-band images in quadrature to produce a detection image in units of sigma (the local standard deviation of noise), which is then searched for local maxima.

The advantages of doing the detection simultaneously at multiple bands are as follows.

• 1.  
  Increased sensitivity to weak sources due to the fact that detection is based on the stack of images at all bands.
• 2.  
  No separate bandmerging step is required, thus avoiding the ambiguities which can occur when trying to associate sources in different bands in the presence of confusion.
• 3.  
  The higher-resolution IRAC data at the shorter wavelengths can guide the extraction at the longer wavelengths (MIPS-24) where the resolution is poorer.

Using the source detection list as an input table of sources to investigate, all measurements are carried out on the combined, mosaic images, and as with the detection step, the extraction is carried out on all bands simultaneously. Integrated fluxes are derived from optimal, inverse-variance weighting point-spread function (PSF) fits to all bands simultaneously. PSF-fitting represents a maximum likelihood estimate of the source position and the set of fluxes at the five Spitzer wavelengths for each source candidate identified by the detection step. The candidate source and its neighbors (i.e., adjacent candidates whose PSF responses overlap significantly with the primary candidate) are grouped into blends, and their parameters estimated simultaneously. Extracted measurements include the equatorial positions and corresponding uncertainties, integrated fluxes from both PSF-fitting and aperture measurements, χ² fit metrics, band-to-band colors and photometric quality flags.

Resolved sources require a separate characterization due to complex shapes and extended emission arising beyond the PSF. Galaxies are identified through visual inspection; their position, shape and size are determined using interactive tools that were designed specifically for 2MASS resolved galaxy characterization (Jarrett et al. 2000, 2003). The integrated flux is extracted from the 1σ elliptical isophote, an adequate proxy for the total flux at the ∼10% level. Extracted measurements include the equatorial positions, elliptical shape parameters of size, axis ratio and position angle, integrated fluxes, mean surface brightness and band-to-band colors.

Low-resolution observations, both IRS-SL and IRS-LL modes, were obtained for both EP-N and EP-S calibration stars to establish the absence of any mid-IR excess above their photosphere fluxes (see Figure 29). The basic SSC pipeline reductions were used to produce a calibrated spectrum for each source, presented in Figure 29 in the Appendix. Although most of these sources have low signal-to-noise ratio (S/N) in the IRS-LL module, they have adequate S/N in the IRS-SL modules (λ< 14 μm) to assess the spectral shape across the W1 and W2 bands of WISE.

In Figure 30 in the Appendix, the IRS spectroscopy of calibrators that are common to both WISE and SAGE is presented, including bright off-pole sources used to calibrate the W3 and W4 bands. Since these sources are significantly brighter than the CVZ calibrators, the spectra have better S/N, especially for the IRS-LL modules. As presented in λ⁴F_λ units, the flat profiles demonstrate the characteristic Rayleigh–Jeans photospheric emission of K–M giants.

Finally, the galaxy NGC 6552 in the EP-N was observed with IRS-LL in two separate AORs (AOR Keys = 27304192 and 27304448) in order to develop this source—unresolved by MIPS and IRS—as a WISE W4 calibrator in the northern CVZ. The LL data were carefully cleaned of bad pixels and combined using the SSC software CUBISM (Smith et al 2007). The spectrum is discussed in detail in the following sections.

As with the Spitzer imaging data reductions (see above), source detection and characterization were carried out using tools developed specifically for the WISE mission by the WSDC. The detection step is similar: both use the deep combined mosaics to identify local maxima, representing candidate point sources, that are further deblended and characterized in the profile-fitting operation.

For source characterization, however, there is one important difference between the respective data reductions. For Spitzer, the photometric measurements were carried out on the combined, deep mosaic images. For WISE, the extraction (profile-fitting) is instead carried out on each individual frame for all bands simultaneously. It thus utilizes an optimal combination of the multi-band imaging data for source photometry, mitigating unruly pixels, artifacts, cosmic radiation hits, source saturation, and angular resolution differences.

The multi-frame, multi-band estimation process represents a departure from the traditional procedure, employed in such software packages as DAOPHOT (Stetson 1987) and SExtractor (Bertin & Arnouts 1996), in which detection and characterization are carried out one band at a time. Another motivation for developing new source extraction algorithms is that currently available packages operate on a single regularly sampled image rather than a set of dithered images. The procedures employed in WISE are optimized for the latter case, correcting for optical distortions and outlier (radiation event) rejection in the individual frames that WISE acquires in orbit.

The EP-S calibrators were supported by an extensive ground-based program of optical classification for objects lacking types from the Michigan Spectral Survey (Houk & Smith-Moore 1988). Ground-based optical spectroscopy of all eight EP-S stars was kindly provided by Dr. M. Bessell of the Australian National University during 2007 November 23–26, using the Double Beam Spectrograph of the Australian National University's 2.3 m telescope. The useful wavelength range was from 3200 to 9000 Å.

Spectral features were compared with the empirical spectra in the MILES database (Sanchez-Blazquez et al. 2006; Cenarro et al. 2007) to estimate stellar parameters. Synthetic Kurucz model spectra were generated with the best-fit parameters for each star and those spectra were treated as stellar templates to be normalized by optical and infrared data (Cohen et al. 1999, 2003).

Optical photometry of our EP-S candidate stars was kindly brokered for us by Dr. D. Kilkenny and was undertaken between 2007 December and 2009 January by Drs. R. Sefako, R. F. van Wyk, and D. Kilkenny, using the 20 inch telescope of the South African Astronomical Observatory (SAAO). The observations were reduced at the SAAO by D. Cooper.

Optical photometry, 2MASS photometry from the 2MASS Point Source Catalog (Skrutskie et al. 2006; Cutri et al. 2003) and mid-infrared photometry from IRAS, MSX, and Spitzer were used to construct SEDs and compared with the model spectra integrated over the RSRs from the various instruments. Ratios of observed in-band fluxes to synthetic in-band fluxes and their associated uncertainties are the central theme of the normalization process. The resulting SEDs with the normalized model spectra are presented in the Appendix (Figures 31 and 32).

The central region of the NEP (R.A. = 18^h00^m00^s, Decl. = +$66^{\rm d}33^{\rm m}38\mbox{$.\!\!^{\mathrm s}$}552$ J2000; l = 96384, b = 29812) contains ten standard calibration stars that are used to cross-calibrate WISE with Spitzer IRAC and MIPS-24 measurements; see Figure 2.

Near the pole center lies the Sy-2 galaxy NGC 6552, which is used to calibrate the red response of WISE (details below). At infrared wavelengths, the brightest source in the EP-N is the planetary nebula NGC 6543 ("Cat's Eye" Nebula)—Figure 3, saturating the signal in the long wavelength bands of Spitzer and WISE. For the Spitzer survey that overlaps with WISE (Figure 2), about 15,000 sources were extracted down to 5σ in at least two bands, including ∼400 galaxies that are resolved by Spitzer-IRAC.

IRAC and MIPS-24 photometry of the EP-N Spitzer–WISE cross-calibration standard stars is presented in Table 3 and their IRS spectroscopy in Figure 29. All of the sources are bright, S/N > 100 for the IRAC measurements, and ranging in S/N between 10 and 50 for the MIPS-24 measurements. For the short-wavelength WISE bands, the EP-N stars are all relatively bright, generally exceeding S/N > 50, while for W4 the stars are too faint. The standards KF06T1, KF06T2, KF06T3, KF03T1, KF03T2, KF03T4, and KF02T3 were removed as W3 standards because of their faintness; KF05T1 was removed in W3 due to potential contamination from a nearby bright star. As discussed in Section 3, due to the paucity of W3 and W4 calibrators in the EP-N, a set of bright K/M giant stars, with known IRAC and MIPS-24 photometry (Engelbracht et al. 2007; Cohen et al. 2003), that are located just outside of both polar CVZs were developed, including optical-infrared spectroscopy and spectral classification (see Figure 30) to augment the photometric calibration of WISE. A summary of the total number of sources per band used for WISE photometric calibration is given in Table 2.

ROSAT, XMM, and Very Large Array observations of the barred disk galaxy NGC 6552 (z = 0.02649) reveal a central strong X-ray and radio source, exhibiting both Sy-1 and Sy-2 emission lines, with the type-2 designation the usual adoption (Condon et al. 1998, 2002; Gioia et al. 2003; Henry et al. 2006; Veron-Cetty & Veron 2006; Shu et al. 2007). For both MIPS-24 and W4, as well as IRS-LL, the host disk is unresolved, best characterized by a point source profile and thus making it a viable photometric calibrator at these wavelengths. Its mid-IR spectrum, Figure 5, is characterized by a steeply rising continuum (due to warm dust grain emission) and a few relatively weak (compared to the continuum) high-excitation lines that arise from the central active galactic nucleus (AGN), none of which is located in the MIPS-24 or W4 bands. To test for variability, multiple-epoch measurements of the galaxy were carried out with both Spitzer and WISE (W4 light curve shown in Figure 6), proving that the photometry is stable at the ∼2% level, at least over the period that the measurements were acquired, nearly eight months for WISE measurements. Interestingly, between day 20 and 40 there is a trend of bright measurements relative to the baseline mean, indicating either real variability or, more likely, the photometry is suffering from artifact contamination due to the close proximity of the infrared-bright NGC 6543 "Cat's Eye" planetary nebula. It is not, however, related to any variation in the WISE spectral response since sources measured in all bands show no variations over time beyond the 1% level (as demonstrated in Figure 10, below).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Spitzer photometry of NGC 6543 and NGC 6552 is presented in Table 4, with measurements acquired using a large, ∼100'' elliptical aperture (major axis, diameter) to capture the total flux for each object, particularly necessary in the shorter wavelength bands in which the source is well resolved. Since the "Cat's Eye" imaging was saturated in the long wavelength bands (IRAC-4 and MIPS-24), the cores were recovered using tools developed by the SSC to rectify bright star saturation. The resultant MIPS-24 photometry, 150 ± 15 Jy, is slightly brighter compared to the IRAS 25 μm flux density, "fnu_z" = 135 ± 12 Jy, extracted using the IRAS Scan Processing and Integration (SCANPI) utility. For both NGC objects, the IRAC measurements have been aperture corrected with factors 0.92, 0.95, 0.82 & 0.78 (IRAC-1, 2, 3, and 4, respectively), to account for the calibration differences between point and extended sources observed by IRAC.¹³

WISE integrated flux measurements of NGC 6552 are presented in Table 5, also using a large aperture to capture both the point source and extended emission in the shorter wavelength bands. (No attempt is made to extract the integrated flux of NGC 6543 due to its complexity and the pixel saturation.) Unlike the IRAC extended source measurements, no additional aperture correction is needed for WISE measurements. The photometric results are graphically displayed in the NGC 6552 spectral energy distribution, Figure 5. With the addition of near-IR measurements from the 2MASS Extended Source Catalog (XSC; Jarrett et al 2003), the SED shape shows the characteristic stellar bump in the near-IR window and the steeply rising continuum at mid-IR wavelengths. The broad-band measurements are consistent with the spectral measurements of IRS-LL (Figure 5), but compared to a Sy-2 model, NGC 6552 appears to have a paucity of PAH emission at 7.7 and 11.3 μm, instead resembling more high-luminosity systems in which the PAH have been depleted or destroyed by the powerful radiation field of the central accretion disk.

The central region of the SEP (R.A. = $06^{\rm h}00^{\rm m}00\mbox{$.\!\!^{\mathrm s}$}00$, decl. = $-66^{\rm d}33^{\rm m}38\mbox{$.\!\!^{\mathrm s}$}552$ J2000; l = 276384, b = − 29811) contains eight standard calibration stars (Table 7), of which seven are used to cross-calibrate WISE with Spitzer IRAC and MIPS-24 measurements; see Figure 4.

In addition to the flux calibrators, the EP-S also contains a source that is used to monitor any ice accumulation on the WISE filter assembly. The "ice monitor" is the source AKARI 060059-663615 (S. Oyabu et al. 2011, in preparation), a probable planetary nebula located in the Large Magellanic Cloud and near the center of the EP-S. AKARI IRC spectroscopic observations were secured by Oyabu et al. on 2006 July 21. The spectrum of particular interest for WISE was taken with the 2.5–5.0 μm grism with a spectral resolution of 120 at 3.6 μm. Oyabu et al. present these data from 3.04 to 3.42 μm, and we have used AKARI photometric measurements to complete a near-IR spectrum that covers the WISE W1 band. Shown in Figure 7 (left), it features the strong PAH emission band at 3.3 μm, used to monitor any water-ice accumulation on the optics. The signature of which would be absorption in the W1 band and a subsequent decrease in signal. The photometry results are also shown in Figure 7 (right), presenting the WISE light-curve photometry for each polar passage over the lifetime of the mission. W1 is stable with a statistical variation of less than 3.0%, and likewise the other bands have an rms scatter of less than 2%–3%. The variation we would expect if this PAH band were to degrade (e.g., with τ > 0.25) is more than twice as large as the actual observed rms. Moreover, as we show below, the W1 instrumental zero point offsets shows no variation or change at the ≪1% level over time (see Figure 10). Hence we conclude that the WISE optics were likely free of any water-ice accumulation throughout the cryogenic mission lifetime.

Zoom In Zoom Out Reset image size

For the Spitzer EP-S survey that overlaps with WISE (Figure 4), about 65,000 total sources were extracted down to 5σ in at least two bands, including ∼500 galaxies that are resolved by Spitzer-IRAC. The total number of sources in the EP-S is significantly larger than that of the EP-N survey for two primary reasons: (1) the EP-S survey is slightly larger in area, and (2) the close proximity of the Large Magellanic Cloud more than doubles the total number of stars along the line of sight.

Photometry of the EP-S Spitzer–WISE cross-calibration standard stars, as well as the "ice monitor" is presented in Table 6. As with the EP-N calibrators, the selected standards are bright, S/N > 100 for the IRAC measurements, and range in S/N between 30 and 100 for the MIPS-24 measurements, but are typically too faint for the WISE W3 and W4 observations. The star 05581475 was removed in all WISE bands due to contamination by a close companion. The standards WOHG642, HD271776, HD270485 were removed as W3/W4 standards because of their faintness. A summary of the total number of sources per band used for WISE photometric calibration is given in Table 2.

Employing calibration standards from both ecliptic polar regions, the photometric performance is graphically shown in Figures 8–10. The first figure, Figure 8, demonstrates how the native raw photometric measurements (DN units) of the calibration stars are related to their predicted flux densities. The predicted flux is derived from the in-band flux of the calibrator spectrum (Figures 31 and 32) measured within the WISE photon-counting RSRs. Tables 7 and 8 list the predicted WISE mags for the Spitzer–WISE cross-calibration sources in the CVZs and off-pole locations, respectively. The ratio DN per mJy shows a flat linear trend across the dynamic range spanned by the calibrator brightnesses. The rms scatter about the mean DN/mJy value is 2.4, 2.7, 4.0 and 6.2% for W1, W2, W3, and W4, respectively, dominated by the uncertainty in the predicted flux density (see below).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The DN versus Jy plot is transformed to the more the familiar magnitude units by first converting the flux density to magnitude units using Fν(iso) (Table 1), and the measured integrated flux to magnitude units using the zero point magnitude.¹⁴ The zero point magnitude (per band) corresponds to the value that produces the statistically zero mean difference between the predicted magnitudes and the measured magnitudes of the calibration stars, as demonstrated in Figure 9.

For first-pass WISE processing, the instrumental zero point magnitudes were defined using the first two months of on-orbit data. As a consequence, there is a slight bias in the zero level (i.e., mag₀ is slightly off) as revealed by the full 7 month complement of data; the ensemble inverse variance-weighted mean and rms in Figure 9 is 0.007 ± 0.025 mag for both W1 and W2, 0.017 ± 0.045 mag for W3, and −0.021 ± 0.057 mag for W4. A small adjustment to the WISE instrumental zero point will be made for the final processing data release (scheduled for mid-2012). The standard star ensemble results suggest that the WISE absolute calibration, tied to that of Spitzer, has rms scatter in the standard calibration stars at the 2.4%, 2.8%, 4.5%, and 5.7% levels for W1, W2, W3, and W4, respectively, and similar to the scatter in Figure 8. The observed scatter arises chiefly from the uncertainty in the predicted WISE flux values, reflecting the inherent errors associated with model fitting to the broad-band SED measurements, as well as the uncertainty in the WISE RSRs. These errors are inflated at the longer mid-IR wavelengths, notably in the W4 window; the red calibrator NGC 6552 has a W4 residual of ∼6% (Figure 10), a difference that appears to be systematic for very red sources as measured using Spitzer–IRS spectra from ultra-luminous infrared galaxies (ULIRGS; see the spectrum of Arp 220 relative to the WISE RSRs, Figure 1). Wright et al. (2010) quote −17% and 9% systematic offsets in the fluxes for W3 and W4 measurements of ULIRGS relative to the standard stars, with the red sources appearing too faint in W3 and too bright in W4, respectively.

Expanding Figure 9 along the time axis, Figure 10 shows the residual flux history for the ensemble of calibration stars. Over the lifetime of the mission, the instrumental zero point magnitude is stable to better than 0.1%; quantitatively, the unweighted sample distribution mean and standard deviation of the mean are −0.003 ± 0.001, −0.001 ± 0.001, 0.022 ± 0.002, and −0.007 ± 0.005 for W1, W2, W3, and W4, respectively. Moreover, time variability is ruled out with individual measurements over the lifetime of the mission (Figure 10); repeated observations of the individual calibrators reveal an rms scatter, typically better than 1% (note the small error bars in Figure 9).

Finally, comparison of WISE with Spitzer for all sources in the polar fields is presented in Figure 11. For the short-wavelength bands of WISE, which closely match those of IRAC, the agreement is very good, within 2%–3%, while for the longer wavelength bands where the filter response is significantly different (e.g., 12 μm versus 8 μm), there is a 15% offset in the sense that W3 and W4 are brighter than IRAC-4 and MIPS-24, respectively. These differences are consistent with the respective RSRs (Figure 1) and the mid-IR Rayleigh–Jeans spectrum of most field stars. At the faint end, W1 and W2 > 16 mag, departures arise due to flux biases associated with low-S/N detection thresholding and source blending/confusion. In the next section we show the source counts for the Spitzer and WISE surveys, noting that at the short wavelengths the depths are comparable.

Zoom In Zoom Out Reset image size

Differential source counts for the Spitzer surveys of the poles are presented in Figure 12. The counts correspond to the area in which full coverage across all Spitzer bands was attained, roughly 0.40 deg² and 1.28 deg² for the EP-N and EP-S, respectively. To improve reliability and avoid spurious detections at the faint end, only sources with at least two-band detections (S/N > 4) are counted. Additionally, the resolved planetary nebula NGC 6543 and its associated emission knots have been carefully removed from the EP-N catalog. The expected Galactic star-counts (adapted from Jarrett et al. 1994) in the L-band (3.5 μm) can be directly compared with the IRAC-1 counts. For the extragalactic contribution, we show the K-band galaxy counts from the 2MASS XSC (Jarrett 2004), extending to K = 14th mag, and the combined Glazebrook et al. (1994) and Cowie et al. (1990) counts that extend to 22nd mag. The K-band magnitude has been adjusted by 0.1 mag (fainter) to predict the L-band galaxy counts, taking into account the typical K−L colors of galaxies with a redshift that is less than one. At the bright end, the EP-N source counts are comprised of Milky Way stars and resolved galaxies, while at the faint end (IRAC-1 > 14 mag) the deep extragalactic sky dominates the counts.

Zoom In Zoom Out Reset image size

For the EP-S, the IRAC-1 source counts are ∼3× larger than the expected Galactic counts, due to the presence of the Large Magellanic Cloud, which dominates the foreground stellar component along the direction of the EP-S.

For the Spitzer EP-N survey, the source counts turn over¹⁵ at 17.5, 17.5, 16.1, 15.9, and 10.5 mag, respectively, for IRAC-1, IRAC-2, IRAC-3, IRAC-4, and MIPS-24, equivalent to 28, 18, 42, 28, and 470 μJy. Although the total area is larger, the depth in the EP-S is somewhat shallower due to the lower coverage depth, the source counts reach a ceiling at 16.4, 16.4, 15.3, 14.5, and 10.6 mag, equivalent to 77, 50, 87, 100, and 430 μJy. As noted above, the EP-S counts are dominated by luminous, evolved stars in the LMC. In both polar fields, the IRAC-2 channel is the most sensitive window to the ecliptic poles. Due to the superior resolution and extraction depth, we will use the Spitzer source counts to estimate the extraction completeness of the WISE counts (see below).

Located just 3 deg from the EP-N is the IRAC Dark Field, used to photocalibrate Spitzer during its six year cryogenic mission. Although the total area is relatively small, co-added IRAC imaging of this Dark Field achieves an extraordinary depth that is only limited by the extragalactic confusion noise (Krick et al. 2009). In Figure 13, a comparison between the Polar survey and that of the Dark Field is presented. To the depth of the EP-N IRAC survey (28, 18, 42, and 28 μJy, respectively), the two surveys agree very well, to a level of ∼2%–4%, implying that the EP-N survey is at least 90% complete to these depths. At the bright end, the relatively small Dark Field likely suffers from small number statistics and consequently a slight deviation is seen in IRAC-2 and IRAC-3. The deep IRAC counts demonstrate the slow roll-off of sources in the EP-N survey, particularly for IRAC-3 and IRAC-4 at depths of ∼30 μJy.

Zoom In Zoom Out Reset image size

The MIPS-24 galaxy counts are compared with those of the Spitzer Wide-area InfraRed Extragalactic (SWIRE) program (Lonsdale et al. 2003, 2004), which covered high-Galactic and ecliptic latitude fields. Following the convention of Shupe et al. (2008), the extragalactic counts are "flattened" by applying a Euclidean scaling: computing the differential counts with each source individually scaled by its flux density to the power of 2.5. For a spatially uniform distribution of sources we would expect a relatively flat trend in log–log space.

The 24 μm counts are shown in Figure 14, where the EP-N is shown in black, the EP-S in gray, and the average-combined SWIRE result in dashed gray. Stars have been removed using color constraints; as shown in the next section, stars separate from galaxies in various combinations of Spitzer colors, as follows (in mag units): [3.6] < 10.5, [3.6] − [5.8] < 0.40 and [4.5] − [8.0] < 1.0, [8.0] − [24] < 2.50 and [4.5] − [8.0] < 0.80.

Zoom In Zoom Out Reset image size

Suffering from small sample fluctuation, the bright end, S(24 μm) > 6 mJy, shows an excess (relative to SWIRE) that is likely due to foreground Milky Way and LMC stars that have not been removed with the color thresholding. Other than the stellar excess at the bright end, the Euclidean-normalized counts of the EP-N and that of SWIRE are flat between 1 and 6 mJy, then steeply rise at the faint end before incompleteness sets in. The rising counts were first discovered with deep ISOCAM 15 μm source counts, attributed to populations at z ∼ 0.8 (arising from the redshifted 7.7 μm PAH) and extending to z = 2 (Franceschini et al. 2001). However, the full extent of this rise in counts is not easily predicted by subsequent galaxy source count models (e.g., see Figure 6 in Shupe et al. 2008), which may indicate that an extragalactic population is significantly under-counted in the modeling. It has been hypothesized that a dust-obscured population (e.g., Type-2 AGN) has been previously undetected or at least under-counted in optically selected studies, but with deep infrared surveys, such as SWIRE and WISE, they are better revealed (Papovich et al. 2004; Shupe et al 2008).

Differential source counts for the WISE survey are presented in Figure 15, and alternative versions in Figures 16 and 17. The total area used for the source counting statistics is 1.54 deg² centered on the pole and extending 0.7 deg, largely overlapping with the Spitzer surveys. Consistent with the Spitzer count methodology, only sources with at least two-band extractions (S/N > 4) are counted, and NGC 6543 and its associated distributed emission has been removed from the EP-N catalog. For comparison, the expected Galactic star-counts (adapted from Jarrett et al. 1994) in the infrared L and M-bands (3.5 and 4.7 μm) can be directly compared with the W1 and W2 counts, while the infrared N and Q-band model counts from the SKY simulation (Cohen 1994) are compared with W3 and W4. For illustrative purposes, we also show the K-band galaxy counts from the 2MASS XSC (Jarrett 2004), Glazebrook et al. (1994) and Cowie et al. (1990) adjusted by 0.1 mag to predict the L-band galaxy counts. At the bright end, W1 brighter than 14th mag, or 0.8 mJy, the EP-N counts (Figure 16) are dominated by Galactic stars (with the brightest end comprised of disk giants), and at the faint end by unresolved galaxies, with the stellar-to-galaxy crossing point at ∼17th mag (∼50 μJy). As with the Spitzer southern survey, the WISE EP-S counts, relative to the expected Galactic star counts, have an excess due to the LMC population. For both poles at the longer wavelengths of 12 and 22 μm, the measured counts are dominated by the extragalactic sky.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The EP-N counts peak at 17.2, 16.9, 13.2, and 9.5 mag, respectively, for W1, W2, W3, and W4, or equivalent to 41, 30, 166, and 1355 μJy, while the EP-S counts peak somewhat brighter at 16.8, 16.6, 13.1, and 9.4 mag, equivalent to 59, 39, 182, and 1450 μJy. The EP-S has less depth (notably in W1 and W2) due to source crowding from LMC stars; and general, compared to Spitzer detection and extraction, the large beam of WISE ultimately limits the depth for detection and extraction.

To estimate the completeness of the WISE catalogs, the deeper and better spatial resolution Spitzer data are employed by comparing the W1 and IRAC-1, W2 and IRAC-2, and W4 and MIPS-24 extractions for the respective EP data sets. The completeness is defined to be ratio of the number of matching WISE-to-Spitzer sources to the total number of Spitzer sources in a given flux bin. A match consists of a spatial association using a 2 arcsec matching radius. Since there is close correspondence between W1:IRAC-1, W2:IRAC-2 and W4:MIPS-24 photometry for S/N > 5 WISE extractions (see Figure 11), the differential completeness with relatively large 0.5 magnitude bins should be accurately tracking WISE relative to Spitzer. Since W3 does not have a close-matching Spitzer equivalent (see Figure 1), it is not as straightforward to estimate the W3 completeness; however, since W3 does not suffer from confusion or blending, it is likely that the completeness is better than 90% where the counts turn over (see W4 results, below).

The differential completeness for both poles is presented in Figure 18. It is immediately apparent that the EP-N is reaching greater depths than the EP-S, a direct outcome of the greater source density in the EP-S increasing source crowding and thereby limiting source detection/extraction. The 90% completeness limits for the EP-N are 16.3, 16.3, and 10.1 mag, respectively, for W1, W2, and W4, or equivalent to 94, 52, and 763 μJy. The EP-N counts peak at lower completeness for the short wavelength bands where confusion is at a maximum; 80%, 78%, and >95% for W1, W2, and W4, respectively. Source blending and confusion effects are the primary reason why the WISE and Spitzer source counts do not match at lower S/N; see Figures 16 and 17, notably seen in the south ecliptic pole. The 90% completeness limits for the EP-S are 14.8, 14.8, and 10.1 mag, respectively, for W1, W2, and W4, or equivalent to 370, 207, and 763 μJy. At the peak in EP-S counts, the completeness has fallen to 65%, 65%, and >95% for W1, W2, and W4, respectively. Note that the long wavelength band, W4, is not suffering from confusion or source blending relative to MIPS-24 because the band is most sensitive to the extragalactic population. Similarly, we believe the interstellar medium (ISM)-sensitive band W3 is also complete to its S/N = 5 limit.

Zoom In Zoom Out Reset image size

WISE provides a unique opportunity to study the 12 μm extragalactic sky at high spatial resolution (compared to IRAS, for example). Using the same convention to study the 24 μm counts (Figure 14), the W3 galaxy counts have been Euclidean-normalized, presented in Figure 19. Stars have been removed using color constraints; as shown in the next section, stars separate from galaxies in various combinations of WISE colors, as follows: [3.4] < 10.5 mag, [4.6]-[12] < 1.2 mag, and [3.5]-[4.6] < 0.50 mag.

Zoom In Zoom Out Reset image size

Between 0.3 and 3 mJy, the counts are flat, similar to that seen in the 24 μm results, and Hacking & Houck (1987) in their deep IRAS survey of the NEP, find a similar number of sources, 25 deg⁻² mag⁻¹ in their faintest flux bin, ∼10 mJy, consisting mostly of foreground stars. WISE extends considerably deeper than IRAS, yet unlike the 24 μm counts at the faint end, there is no sign of a steep rise in the counts, suggesting that the WISE 12 μm channel is not as sensitive to galaxies at higher redshift. Similarly, the shorter-wavelength IRAC-4 (8 μm) is also flat at the faint end. A more detailed analysis of WISE 12 μm counts in the NEP and in the Bootes field direction will be presented in D. Benford et al. (2011, in preparation).

Color–magnitude diagrams are presented in Figure 20. Resolved galaxies (green circles) are distinguished from point sources (filled circles), as is the Seyfert galaxy NGC 6552 (blue circle) and planetary nebula AKARI 060059-663615 (magenta square). Stars generally fall near zero color and span a wide-range in flux, with luminous supergiants representing the brightest sources. In the EP-S, LMC stars comprise the bulk of the diagram, separating into evolutionary main-sequence and giant tracks. As detailed in Blum et al. (2006) and Bonanos et al. (2010) study of SAGE colors, red supergiants (RSGs) are the most luminous [3.6] sources, while blue supergiants, sgB[e], have the reddest [3.6]-[8.0] and [8.0]-[24] stellar colors. The coldest Galactic object in the field is the planetary nebula, whose dust-shell color temperature is likely to be ∼160 K based on its mid-IR colors closely matching those of PN NGC 1514 (Ressler et al. 2010).

Zoom In Zoom Out Reset image size

The faintest (and most distant) sources are extragalactic, which trend toward redder colors due to the primary contribution from stellar (evolved stars), ISM (dust re-emission arising from star formation) and AGNs. However, early-type galaxies, with their stellar-dominated emission, may have overlapping colors with foreground stars. The brightest galaxies are usually resolved by Spitzer, indicative of their relative close proximity. Most notably, NGC 6552 is the brightest source in the EP-N at long wavelengths, owing to its nearby location (z = 0.026) and steeply rising AGN-dominated continuum (Figure 5).

Photometric colors separate not only most stars from galaxies, but also AGN-dominated galaxies from normal and star-forming galaxies, as demonstrated in Figures 21 and 22. Studied in detail by Lacy et al. (2004; 2007) in Figure 21, and Stern et al. (2005) in Figure 22, the red-dashed boxes denote the regions of color–color space that AGNs—both Type-1 and Type-2—tend to occupy. This mid-IR color separation arises from the power-law, F_ν ∼ ν^−α, dominated emission from the central AGN compared to normal or star-forming galaxies that have a distinctive stellar Rayleigh–Jeans fall-off, followed by strong PAH emission and rising continuum due to warm dust re-emission associated with star formation. Moreover, the AGN-dominated galaxies tend to be at higher redshift (z ∼1–3) compared to field galaxies (z < 0.6; Stern et al. 2005). Note that NGC 6552, a nearby, low-luminosity Type-2 AGN, falls just blueward ([3.6]-[4.5]) of the Stern et al. (2005) box, not uncommon with obscured, low-luminosity AGNs and emission line galaxies (see also Eckart et al. 2010; Donley et al. 2008; Seymour et al. 2007); it does, however, fall within the Lacy et al. (2004) box. In Figure 22, low-redshift star-forming galaxies tend to have red [5.8]-[8.0] and blue [3.6]-[4.5] colors, occupying the lower right portion of the diagram, while ellipticals and early-type spirals migrate with redshift from the central locus to the upper-left side of the diagram (see Eckart et al. 2010).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Both Type-1 and 2 AGNs represent a relatively large fraction of the total extragalactic population comprising the Spitzer study of the ecliptic poles. After removing foreground stars (see Section 5.3.2 below), and limiting IRAC-4 to a flux limit of 75 μJy, the ecliptic pole AGN box incorporates a source density of 318 deg⁻² in the EP-N, and 230 deg⁻² in the EP-S, which is comparable to the IRAC Shallow Survey density of AGNs, 275 deg⁻² (Stern et al. 2005) to the same depth. The observed lower density in the EP-S is a consequence of the relatively incomplete (across the field) depth of the EP-S IRAC observations compared to the uniform EP-N and the IRAC Shallow Survey (Eisenhardt et al. 2004).

Finally, the longest wavelength colors are presented in Figures 23–25. The first two are similar to the colors that WISE produces (see below), while the last highlights the coldest colors that the Spitzer mid-IR survey provides (note however, we are not presenting any 70 or 160 μm colors). Stars tend to cluster in the lower left side of the diagram with their relatively blue (hot) colors, early-type galaxies with their luminous evolved population overlap with stars at the redder end of the stellar locus, while star-forming galaxies cluster in the upper right, fully spanning the red (cool) regions of the diagram.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

An interesting feature of the EP-N diagram is the clear separation of resolved galaxies with the more distant unresolved extragalactic population, whose colors tend to be redder in [8.0]-[24] color due to enhanced mid-IR continuum arising from star formation, the presence of AGNs and emission features shifting into the long wavelength channel (e.g., the strong 11.3 μm PAH band observed in starburst galaxies shifts into the 24 μm window at z ∼ 1). This distant population is not as pronounced in the EP-S because of the shallower depth, sensitive to only lower-redshift galaxies. Moreover, the EP-S has more unresolved sources with redder [4.5]-[8.0] colors, presumably from the LMC population of supergiants, planetary nebulae (H₂ and/or PAH emission dominates the broad 8 μm band) and young stellar objects (YSOs).

Combining the color–magnitude and color–color diagrams for both EP surveys, stars reliably separate from galaxies using the following Spitzer IRAC/MIPS apparent brightness and color criteria:

[3.6] < 10 mag; [24] < 5 mag; [3.6]-[8.0] < 0.80 mag; [8.0]-[24] < 1.0 mag; [3.6]-[5.8] < 0.4 and [4.5]-[8.0] < 1.0 mag; [8.0]-[24] < 2.5; and [4.5]-[8.0] < 0.80 mag.

These thresholds are used to remove stars from the Spitzer catalog in order to construct the extragalactic 24 μm counts (see Figure 14).

WISE color diagrams have interesting and subtle differences from those of Spitzer due to the W3 (12 μm) band. The W3 bandpass is sensitive to PAH (11.3 μm) emission from nearby galaxies and shorter wavelength PAH (6.2 and 7.7 μm) emission from redshifted galaxies, as well as warm continuum from AGNs at both low and high redshift (see Figure 1 (right)). Showcasing the WISE band at 12 μm, the EP survey colors (S/N > 5) are presented in Figure 26. For illustrative and interpretation purposes, the predicted WISE colors for extragalactic populations based on templates from SWIRE augmented by GRASIL models (Polletta et al. 2006, 2007; Silva et al. 1998; see also Wright et al. 2010) are depicted as follows: the shaded regions of different galaxy types denote where the synthesized photometry of redshifted sample galaxies, using simulated SEDs, drops out from the WISE detection (where we have adopted a modest WISE sky coverage of ∼12 passes). For example, Type-1 luminous QSOs occupy a region of color–color space denoted in light blue; their colors range with redshift out to z ∼ 2 where they become too faint for WISE to detect. Normal ellipticals and spirals are seen out to z ∼ 0.5, while luminous infrared galaxies (LIRGs) and low-ionization nuclear emission-line regions (LINERs) are seen to z ∼ 0.8 and Seyferts to z ∼ 1, respectively.

Zoom In Zoom Out Reset image size

A key feature of these color–color diagrams is that galaxies clearly separate from field stars owing to their redder colors; nonetheless, a wide variety of galaxy colors belies the rich and diverse extragalactic sky. Stars have colors near zero magnitude, trending slightly redward in [4.6]-[12] color with the more luminous, evolved populations. Early-type field galaxies occupy the "green" zone, extending toward redder colors and disks that are actively forming stars. The most extreme star-forming galaxies, LIRGs (tan/brown), starbursts (yellow) and ULIRGs (orange) are relatively rare; only a small fraction of the total are evident in the EP diagram. For the EP-S, YSOs forming in the LMC have colors that are similar to those of obscured galaxies and are likely to be confused as such. Finally, AGNs are clearly present in the WISE survey, extending from the field spirals toward redder colors in [3.4]-[4.6] due to their power-law mid-IR spectrum, occupying the QSO (blue), Seyferts (cyan), and obscured AGN (green) regions of the diagram. In the following analysis, we attempt to quantify the number of AGNs observed by WISE in the EP surveys using the model-delineated diagram, Figure 26, comparing with color–color methods developed for IRAC photometry of extragalactic fields (see Section 5.3.1).

We first define a general "AGN" region, encompassing QSOs and Seyfert galaxies. To avoid degeneracy with field galaxies that set in near the blue end of the [3.4]-[4.6] AGN color, we restrict the AGN "box" to lie above where most of the field galaxies occupy in the [3.4]-[4.6] color of Figure 26. The empirical criteria are thus defined to be (see dashed gray box Figure 26):

$$\begin{eqnarray} &&[4.6]-[12] > 2.2 \rm \, mag \ \rm and \ [4.6]-[12] > 4.2 \rm \,mag,\nonumber \\ &&[3.4]-[4.6] > (0.1\,\times \,[4.6-12] + 0.38)\, \rm mag\nonumber \\ &&\rm and\ [3.4]-[4.6] < 1.7\, \rm mag. \end{eqnarray} \tag{ 1 }$$

It is inevitable that some normal galaxies will scatter into the AGN box, and conversely some AGNs will scatter out of the box, in addition to the bluer (i.e., host-dominated) AGNs that lie outside of the defined region (e.g., NGC 6552).

The next step towards estimating the number of AGNs in the WISE EP surveys is to consider the contaminating contribution from stars, with the goal of removing as many foreground and LMC stars from the diagram and subsequent analysis as possible given the information at hand. For this task, we not only use the WISE color–mag and color–color information presented thus far, but also the color that the longest wavelength band provides: the [12]-[22] color (Figure 27), cleanly isolates the foreground stars from the extragalactic population, similar to what is observed with the IRAC-MIPS-24 colors (Figure 25). Combining the various diagrams, stars separate from galaxies with the following empirical WISE-only criteria: [3.4] < 10.5 mag, [4.6]-[12] < 1.5 and [3.4]-[4.6] < 0.4 mag. For [12]-[22] color, galaxies are redder than 1.2 magnitude and stars bluer than this limit. Galactic objects with extreme colors (e.g., blue supergiants, planetary nebulae, and YSOs) may lie outside of these limits, mixing with the extragalactic population; however, beyond the Galactic plane and the LMC (EP-S) they are relatively rare and will be statistically insignificant, as is the case for the EP-N.

Zoom In Zoom Out Reset image size

After removing stars from the EP sample using these WISE color limits, the WISE AGN box now encompasses a source density of ∼260 deg⁻² in the EP-N and 160 deg⁻² in the EP-S down to a W3 limit of 115 μJy, which is roughly 15% of the total number of extragalactic sources, comparable to the AGN fraction seen in low luminosity systems (e.g., Goto et al. 2011). The lower fraction observed in the EP-S is likely due to confusion (and corresponding loss in completeness, see Figure 18) from the higher stellar density arising from the foreground LMC.

Mapping the AGN-selected WISE sources back to the corresponding Spitzer measurements, the resulting diagram (Figure 28) reveals where WISE-selected AGNs are located in the Spitzer color–color plot. WISE AGNs are denoted with filled, blue points: for the most part, WISE AGNs are located within the Stern et al. (2005) box, thus validating the WISE-determined AGN box. For those outside of the box, the colors tend to be slightly blue in [5.8]-[8.0] color, the region of the diagram where higher redshift early-type galaxies (which tend to have active nuclei) tend to occupy, which suggests that the Stern et al. (2005) AGN box may be expanded to a value of ∼0.4 mag in that color. A more detailed analysis of WISE colors and AGNs in the Cosmological Evolution Survey (COSMOS) field is presented in D. Stern et al. (2011, in preparation), and in L. Yan et al. (2011, in preparation) and E. Donoso et al. (2011, in preparation) who are using SDSS data to study galaxy populations.

Zoom In Zoom Out Reset image size