The Spitzer Coverage of HSC-Deep with IRAC for Z studies (SHIRAZ). I. IRAC Mosaics

In Cycle 14, we were awarded 488.33 hr on the Spitzer Space Telescope to extend the Spitzer IRAC coverage in E-COSMOS, Deep2-F3, and EN1 (PID:14081; PI: A. Sajina). The observing strategy followed that of the SERVS (Mauduit et al. 2012) and Spitzer DeepDrill surveys (Lacy et al. 2021), with six small cycling dithers per pointing, using 100 s frametime. These Astronomical Observation Requests or AORs tile the desired areas with small overlaps to minimize gaps given the potential differences in relative orientation. The AORs are grouped into two observing epochs, separated by ≈6 months, which helps remove any moving objects such as asteroids. In addition, the observations of the two channels are carried out simultaneously but are slightly offset on the sky. Every six months, the orientation of the telescope rotates by 180°, effectively flipping the relative orientation of the two channels. Thus the two epochs help maximize the area of overlap between the two channels. A summary of the new observations in these fields is presented in Table 1. This summary includes the number of AORs per field and per epoch.

In this section, we discuss the archival Spitzer IRAC data that have been combined with our new observations to produce our final IRAC mosaics.

• 1.  
  DEEP2-FIELD3 (DEEP2-F3) is a 4 deg² field at R.A. = 23^h, decl. = −00^d. This field was previously mapped by SpIES (Timlin et al. 2016), which extended over about 100 deg² of the SDSS Stripe82 field (PID 90045; PI G. Richards). These data, however, are shallow with only 60 s exposures per pointing. These data are so shallow that we decided to not combine them with our observations.
• 2.  
  E-COSMOS is a 5 deg² field at R.A. = 10^h, decl. = 2^d whose central 2 deg² covers the original COSMOS field (Scoville et al. 2007). COSMOS has previous IRAC data coverage coming from multiple surveys, e.g., S-COSMOS (Sanders et al. 2007), the Spitzer Extended Deep Survey (Ashby et al. 2013, SEDS), S-CANDELS, (Ashby et al. 2015), Star Formation at 4 < z < 6 from the Spitzer Large Area Survey with Hyper Suprime-Cam (SPLASH; Steinhardt et al. 2014), and Spitzer Matching survey of the UltraVISTA ultradeep Stripes (SMUVS; Ashby et al. 2018).
• 3.  
  ELAIS (European Large Area ISO Survey) North 1 (EN1) is a ∼8 deg² field at R.A. = 16^h decl. = +55^d that was previously imaged in its entirely in both 3.6 and 4.5μm as part of the Spitzer Wide-area InfraRed Extragalactic survey (SWIRE; Lonsdale et al. 2003). However these data are shallow with only 80 s exposures per pointing. A subsection of 2 deg² of this field has exposures of 1200 s per pixel from the Spitzer Extragalactic Representative Volume Survey (SERVS; Mauduit et al. 2012). Our combined mosaic for EN1 includes both SERVS and SWIRE in addition to the new observations described in Table 1. The SERVS data in EN1 were obtained early in the Spitzer warm mission and subsequently the absolute calibration of the instrument was adjusted. To correct for this calibration difference, multiplicative factors of 1.041 in the 3.6μm channel and 0.980 in the 4.5μm channel were applied to the SERVS data before including them in the combined mosaics as described in the following sections (see also Lacy et al. 2021).

In Table 2, we list all the programs that have been used to derive the master mosaics in COSMOS and EN1.

Figure 1 shows the IRAC coverage maps of all four HSC-Deep fields. For completeness, this includes XMM-LSS, which was observed as part of the Spitzer DeepDrill survey (Lacy et al. 2021). The coverage maps for the other three fields are discussed in more detail in the subsequent sections of this paper. For all fields, we overlay the most crucial ancilary data including the HSC-Deep, CLAUDS, and near-IR surveys. The later include the VIDEO survey (Jarvis et al. 2013), the VEILS survey, and the UKIDSS survey (Lawrence et al. 2007). It can be seen that with the addition of our new IRAC observations, we now have deep IRAC coverage to go with the areas of overlapping U through near-IR coverage. Note that in this paper we only present the IRAC mosaics for EN1, Deep2-F3, and E-COSMOS. The mosaic for XMM-LSS is presented in Lacy et al. (2021). Figure 2, shows the area and sensitivity of our mosaics (see Section 3.2.2) in the context of other surveys at 3.6 μm.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We downloaded the Spitzer pipeline processed data for all AORs (see Table 1) observed within this program as well as from archival programs in these fields from the Spitzer Heritage Archive. All AORs observations are reduced to basic calibrated data (cBCD), using the Spitzer Science Center (SSC) calibration pipeline. In this step, the SSC subtracts dark current, performs detector linearization and flat-fielding, corrects for artifacts like column pull-ups and pull-downs, multiplexer bleed, first frame effect etc., Then it provides uncertainty estimates, bad pixel mask, and cosmic-ray rejection masks for each frame. Moreover, the images are absolutely calibrated into physical units (i.e., MJy sr⁻¹ = 10⁻¹⁷erg s⁻¹ cm⁻² Hz⁻¹ sr⁻¹).

Then, we use the reduction pipeline described in Labbé et al. (2015), Annunziatella et al. (2018) and Stefanon et al. (2021) to postprocessed the cBCDs. Basic calibrated data from different programs were generated using different SSC pipeline versions over the years. This particularly affects astrometry, image distortion refinements, and artifact correction. By postprocessing all cBCDs with the same reduction pipeline, we overcame these issues in our final mosaics. The reduction process consists in a two- pass procedure, where each AOR is reduced independently. The first pass includes: initial background and bias structure estimation from a median of all the frames in the AOR and the consequent subtraction; the correction of column pull-up and pull-down introduced by cosmic rays and bright stars; and persistence masking and muxbleed correction rejecting all highly exposed pixels in the subsequent four frames.

The second pass includes cosmic-ray rejection, astrometric calibration, and an accurate large-scale background removal. Cosmic rays are removed by using an iterative sigma clipping method. The background level is first estimated as the median of the frames in each AOR masking sources and outlier pixels. Then, it is refined by clipping pixels associated to objects and subtracting the mode of the background pixels. Single background-subtracted frames are then combined into a mosaic (using WCS astrometry to align the images). In this step, we masked bad pixels and applied a distortion correction in WCS. This image, in combination with the reference image, is then used to refine the pointing of the individual mosaics. The individual pointing-refined frames are registered to and projected on the reference image. The reference images for the three fields are the z-band images from the HSC-Deep survey from the second public release (pdr2). For these images, the astrometric calibration was carried out against the Pan-STARRS1 DR1 catalog (Chambers et al. 2016; Aihara et al. 2019). The pdr2 website does not release the entire mosaic of each field, but each of them has been divided in tracts of 1.7 × 1.7 deg² (Aihara et al. 2019), and each tract in turn into 81 patches of $\sim \ 12^{\prime} \times \ 12^{\prime}$. We downloaded the patches related to each field from the website and combined them in tracts using the python code available on the website. Then, we combined all the tracts of a field using SWarp (Bertin et al. 2002), while resampling them to a pixel scale of 06 pixel⁻¹ (0.5×the IRAC original pixel scale). For each field, we produce different end-type images:

• 1.  
  Single-epoch mosaics;
• 2.  
  Epoch 1 and Epoch 2 combined mosaic;
• 3.  
  Combined data of all the observations available in that field.

While we made a master mosaic of E-COSMOS that includes all of IRAC data, most of our analysis is based on the combination of S-COSMOS plus our new observations which pad out the "corners" of the classical COSMOS field. We chose to do so in order to keep to a relatively uniform depth.

The left-hand side of Figure 3 shows for illustration the final IRAC EP1+EP2 mosaics relative to the reference HSC z-band image. This figure illustrates the good overlap between the channel1 and channel2 coverage of each field thanks to the AORs being split into 2 epochs 6 months apart. The right-hand side of Figure 3 shows the coverage maps in this field based only on our new data. These both illustrate the observing strategy as well as the highly uniform coverage in DEEP2-F3 where the mosaic is essentially all our own data. By contrast, Figure 4 shows the more variable exposure in E-COSMOS (especially the deeper data in the central 2 deg²) and EN1 (where the shallower outskirts are the SWIRE data, whereas the deeper irregular shaped center is SERVS+our data).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In this section, we describe the quality of the final mosaics in both channels for each of the three fields presented in the previous section. To this end, we use all of the available data in EN1 and DEEP2-F3, and only the data corresponding to this program and to S-COSMOS for the E-COSMOS field.

3.2.1. Point-spread Function

To characterize the full width half maximum (FWHM) of the point-spread function (PSF) of these mosaics we define a sample of ∼100 bright, isolated and unsaturated point sources in each channel and each field. We identified the stars from the SExtractor catalogs in each filter and each field, by combining several methods described in Annunziatella et al. (2013) and by visually verifying them. This sample is used to derive the FWHMs of the IRAC images. The derived FWHM are listed in Table 3. The FWHM varies from 169 in EN1 to 177 in DEEP2-F3 at 3.6 μm, while it varies from 160 in EN1 to 167 at 4.5 μm. We also used this sample to construct stacked PSF images in each field and each channel. These images showed a fairly stable PSF across the fields. For a more quantitative assessment, Figure 5 shows the curves of growth across each field (and for all fields combined) where the top panels are for the 3.6 μm channel whereas the bottom panels are for the 4.5 μm channel. Each curve is normalized to the flux in a 6 '' aperture. Such curves of growth were constructed for each of our ∼100 unsaturated bright point sources. Figure 5 also shows relatively little spread among our sample of bright unsaturated stars, consistent with a stable PSF across the mosaics. We use the median growth curve in each field to derive the median half-light and 75%-light radii, i.e., the radii that contain 50% and 75% of the light, respectively. These values are reported in Table 3, together with their 1σ uncertainties. We checked for possible variations of the growth curves when using EP1, EP2, EP1+EP2, or combined mosaics, and we did not find any. For this reason, we only report the result obtained when using the combined mosaics (e.g., EP1+EP2 for DEEP2-F3, EP1+EP2+S-COSMOS for E-COSMOS, and EP1+EP2+SWIRE+SERVS for EN1).

Zoom In Zoom Out Reset image size

Table 3. Characteristics of the IRAC Observations

Filter	FWHM	50% Light Radius	75% Light Radius
(μm)	('')	('')	('')
E-COSMOS:
3.6	${1.72}_{-0.10}^{+0.14}$	${1.15}_{-0.04}^{+0.04}$	${1.80}_{-0.06}^{+0.07}$
4.5	${1.61}_{-0.08}^{+0.10}$	${1.17}_{-0.04}^{+0.04}$	${1.91}_{-0.04}^{+0.03}$
DEEP2-F3:
3.6	${1.77}_{-0.08}^{+0.14}$	${1.15}_{-0.04}^{+0.05}$	${1.80}_{-0.04}^{+0.05}$
4.5	${1.67}_{-0.11}^{+0.12}$	${1.15}_{-0.05}^{+0.03}$	${1.90}_{-0.07}^{+0.04}$
EN1:
3.6	${1.69}_{-0.18}^{+0.22}$	${1.21}_{-0.06}^{+0.05}$	${1.86}_{-0.05}^{+0.04}$
4.5	${1.60}_{-0.10}^{+0.12}$	${1.18}_{-0.07}^{+0.07}$	${1.93}_{-0.04}^{+0.03}$

Download table as:  ASCIITypeset image

3.2.2. Depth

To estimate the depth of each mosaic, we use the "empty aperture" method to empirically determine the noise properties of our IRAC mosaics, following the same approach as in Annunziatella et al. (2018). Briefly, we randomly place a large number of apertures on the noise-normalized images (obtained by multiplying the images by the square root of the coverage maps), reject all apertures falling on sources (identified by SExtractor; see Section 4), and measure the flux in the remaining ones. We chose an aperture size of 3'' and measure the flux in ∼4000 apertures in each field and each channel. A histogram of the measured fluxes is constructed for each field in both 3.6 μm and 4.5 μm. These histograms are all well fitted by Gaussians. The best-fit widths (σ_best−fit) of these histograms can be converted to 3σ magnitude depths derived using the empty aperture method according to

$$\begin{eqnarray}&&\mathrm{depth}(3\sigma )[{AB}]=-2.5\mathrm{log}\left(\displaystyle \frac{3{\sigma }_{\mathrm{best}-\mathrm{fit}}}{\sqrt{\mathrm{COVERAGE}\,\ \mathrm{MAP}}}\right)+{ZP},\end{eqnarray} \tag{ 1 }$$

where ZP is 20.04 for all images. Note that here the division by the square root of the coverage map (aka the exposure time per pixel) reverses the procedure above when we generated noise-normalized images.

Figure 6 shows the 3σ magnitude depth for each field and each channel. As expected, regions with higher exposure times (see Figures 3 and 4) also have fainter 3σ magnitude depths. Table 4 lists the 15th, 50th, and 75th percentiles of the 3σ magnitude depths in both 3.6 and 4.5 μm in an aperture of D = 3'' as derived with the empty aperture method. As expected the depth is shallowest for the EN1 field, channel 2, where we have the lowest exposure time (see Table 1). We also converted these to the more conventional 5σ values and found that on average the 5σ depth is 23.7 at 3.6 μm and 23.3 at 4.5 μm. This is comparable to the depths of the SERVS (Mauduit et al. 2012) and Spitzer DeepDrill (Lacy et al. 2021) surveys.

Zoom In Zoom Out Reset image size

Table 4. Depths of the IRAC Observations

Filter	15th Percentile of 3σ Depth a	Median of 3σ Depth a b	85th Percentile of 3σ Depth a
(μm)	(AB)	(AB)	(AB)
E-COSMOS:
3.6	23.82	24.30	24.49
4.5	23.48	23.84	23.97
DEEP2-F3:
3.6	23.72	24.01	24.16
4.5	23.39	23.74	23.89
EN1:
3.6	22.94	23.54	24.30
4.5	22.44	23.14	24.05

Notes.

^a Using a 3'' circular aperture diameter. ^b To convert these to 5σ simply subtract 0.22 from these numbers. The average 5σ depth for the three fields is 23.7 at at 3.6 μm and 23.3 at 4.5 μm.

Download table as:  ASCIITypeset image

We matched the obtained catalogs to Gaia Data Release 3 (DR3; Lindegren et al. 2018; Torra et al. 2021). The IRAC pointing is calibrated using the HSC-z-band reference image. We matched the position of the sources in our catalogs with those from Gaia DR3 using a 10 match radius. Around ∼4% of the sources in the three fields have counterparts in Gaia DR3. The results are shown in Table 6, where we list the median systematic offset between Spitzer and Gaia DR3 positions (R.A.) and (decl.), along with the scatter σ (R.A.) and σ (decl.), representing the positional accuracy of a typical source in our fields. All systematic offsets are < 01.

Figure 7 shows the comparison between the total magnitudes for the objects detected in EN1 and those from the SERVS catalog (Mauduit et al. 2012) in both the 3.6 μm mosaic (left-hand panel) and the 4.5 μm mosaic (right-hand panel). Our mosaic in EN1 in the area where it overlaps with SERVS is entirely made up of SERVS data (both our mosaic and SERVS coadd the SWIRE data here). Therefore, the only differences are in the data processing and the specific SExtractor settings.

Zoom In Zoom Out Reset image size

In order to estimate the total magnitude for the sources in EN1, we multiply the fluxes of the sources in an aperture of 39 diameter for an aperture correction factor derived using the median growth curves shown in the third panel of Figure 5. The total fluxes for the sources in SERVS are provided in the SERVS catalog and are derived using a similar approach as in this work, but with a slightest smaller aperture (D = 38). While unsurprisingly there are some outliers, the vast majority of the sources show excellent agreement between the earlier SERVS catalog (Mauduit et al. 2012) and our new SExtractor photometry based on a reprocessed EN1 mosaic. Figure 7 shows this comparison in both channels. The comparison is carried out up to AB = 24 mag. There are median offsets between the magnitudes which are: 0.04 at 3.6 μm and 0.05 for 4.5 μm. Some of this is due to our accounting for the recalibration of the 1st year Spitzer warm mission data (see Section 2.2), which was not known at the time of the original SERVS catalog. This level offsets are also not unexpected given the somewhat different SExtractor settings we use relative to the ones in Mauduit et al. (2012). In the figure, there are overplotted in gray the running medians and the correspondent 1σ confidence interval. As it can be seen, there is no trend with magnitude over the entire range.

Figure 8 shows the differential source counts in the E-COSMOS, DEEP2-F3, and EN1 fields measured in 3.6 μm and 4.5 μm. The number counts are derived using aperture magnitudes computed within an aperture of diameter 3.9 scaled to total using the median growth curves shown in Figure 5. The effective area in each field is evaluated by excluding the regions with lower exposure times and shallower depths (i.e., the SWIRE region in EN1, or regions where epochs do not overlap). This leads to an effective area of: 4.41, 2.93, and 3.68 deg² for E-COSMOS, DEEP2-F3, and EN1, respectively. We also overplot the values of the 5σ depth in each field and each channel scaled to total magnitude using the same approach as above. We compared our differential number counts with those from SERVS and S-CANDELS. In both cases the total magnitudes were obtained by applying a correction function to the fluxes derived in apertures of 3.8 and 2.4 diameter, respectively, for SERVS and S-CANDELS. As we can see from Figure 8, the number of detected sources in both channels in EN1 and E-COSMOS is comparable to that of previous surveys, up to the respective 5σ limit.

Zoom In Zoom Out Reset image size

Lastly, we note that we do not remove stars from our counts. The slight upturn seen in the counts at magnitudes ≈18 is due to the transition from the regime where stars dominate the counts to where galaxies dominate the counts (see e.g., Ashby et al. 2015; Lacy et al. 2021).

Accompanying this paper, we present the public data release, consisting of reduced images of all IRAC observations in three fields, namely E-COSMOS, DEEP2-F3, and EN1. All data products will be available on the NASA/IPAC Infrared Science Archive. The data release contains the following:

• 1.  
  Spitzer/IRAC mosaic science images in both 3.6 and 4.5 μm. The images are astrometrically registered to the reference z-band image from the HSC Subaru Public Data release 2.
• 2.  
  Coverage maps containing exposure times (seconds) in both 3.6 μm and 4.5 μm.
• 3.  
  Maps with 3σ magnitude depth (as derived in Section 3.2) at the same WCS and pixel scales of the scientific images.

The primary goal for obtaining these new data and constructing these mosaics combining all available data was to provide IRAC 3.6 and 4.5 μm coverage of the HSC-Deep fields. The mosaics presented in this paper represent a total of 17.9 deg², which together with the XMM-LSS IRAC coverage through SERVS+Spitzer DeepDrill (Lacy et al. 2021) means that the bulk of the 27 deg² HSC-Deep fields now has sufficiently deep IRAC coverage to study massive galaxies out to z ∼ 6 as well as reach well below the knee of stellar mass function at cosmic noon. Indeed, adding the IRAC data to the wealth of U through near-IR photometry already obtained in these areas will lead to improved photometric redshifts but even more significant improvements in the derived stellar population parameters such as stellar mass, age, and star-formation rate (see e.g., Muzzin et al. 2009). This will impact a wide range of galaxy evolution studies to be performed with these data. The wide area of the HSC-Deep survey makes it well suited for the study of the role of environment in galaxy assembly and quenching since it will sample the full range of environments with high statistical significance. For example, in only 4.5 deg² of the XMM-LSS, Krefting et al. (2020) find nearly 400 overdensities consistent with dark matter halos of M > 10^13.7 M_⊙ including several structures that look like cluster mergers and large filamentary structures. With 5.5×the area covered in the present data, the potential to explore galaxy evolution in the context of both local and large-scale environment is great. These data are also well suited to AGN studies including BH-galaxy coevolution studies. For BH-galaxy coevolution studies what is critical is that the combined multiwavelength data will allow for both photometric redshifts and stellar population parameters of all potential AGN host galaxies. The IRAC data are critical for accurate estimates of such parameters as described above. This deep IRAC survey is also potentially useful for finding AGN in dwarf galaxies which may be powered by IMBHs (Satyapal et al. 2017).

Lastly, the HSC-Deep fields will be the sites of the upcoming Prime Focus Spectrograph (PFS) galaxy evolution survey (Tanaka et al. 2017). The PFS is a highly-multiplexed spectrograph with ≈2500 fibers being build for the 10 m Subaru telescope. The PFS spectra will cover λ ∼ 0.3–1.3 μm with R ∼ 3000 providing a wealth of information for the targeted galaxies including nebular-line based SFRs, metallicites, AGN content, and outflow rates to name a few. The multiband photometric coverage of these fields, including IRAC, will allow for the optimal target selection for the PFS as well as providing significant value to the spectroscopic data by allowing for key photometrically derived parameters such as stellar mass.