The Great Observatories All-Sky LIRG Survey: Herschel Image Atlas and Aperture Photometry^*

The Photoconductor Array Camera and Spectrometer (PACS, Poglitsch et al. 2010) is one of three FIR instruments onboard the Herschel Space Observatory and covers a wavelength range between 60 and 210 μm. In the photometer mode, it can image two simultaneous wavelength bands centered at 160 μm, and at either 70 μm or 100 μm. These three broad bands are referred to as the blue channel (60–85 μm), green channel (85–130 μm) and red channel (130–210 μm). For any given observation, the blue camera observes at either 70 μm or 100 μm, whereas the red camera only observes at 160 μm. A dichroic beam-splitter with a designed transition wavelength of 130 μm directs the incoming light into the blue and red cameras, and a filter in front of the blue camera selects either the blue or green band.

The detectors for both the blue and red cameras comprise a filled bolometer array of square pixels that instantaneously samples the entire beam from the telescope's optics. The layout of the blue camera's focal plane consists of 4 × 2 subarrays, with 16 × 16 pixels in each subarray. Similarly, the red camera consists of 2 × 1 subarrays with 16 × 16 pixels each. On the sky, each bolometer pixel subtends an angle of 32 × 32 and 64 × 64 for the blue and red cameras, respectively. There exists gaps between each of the subarrays in both cameras, which must be filled in by on-sky mapping techniques (i.e., scan mapping). Both the blue and red cameras were designed to image the same 35 × 175 field of view on the sky at any given instant.

In the photometer mode, there are two astronomical observing templates available, in addition to a PACS/SPIRE parallel observing mode. For our Herschel GOALS program, we used the scan map technique for all of our astronomical observing requests (AOR), which is ideal for mapping large areas of the sky and/or targets where extended flux may be present. Our scan map observations involve slewing the telescope at constant speed along parallel lines separated by 15'' from each other, perpendicular to the scan direction. Two example PACS observation footprints are shown in Figure 2 panels (a) and (c), overlaid on images from the Digital Sky Survey (DSS). The area of maximum coverage is the inner region centered on the red box, where the requested observation is centered. For the GOALS observations, we chose to observe seven scan legs in each scan and cross-scan using the 20'' s⁻¹ scan speed, with scan leg lengths ranging between 3 and 6' depending on the size of the target. At this scan speed, the beam profiles for each wavelength have mean FWHM values of 56, 68, and 113 for the 70, 100, and 160 μm channels, respectively.

Zoom In Zoom Out Reset image size

Before each PACS photometer observation is a 30 s chopped calibration measurement between two internal calibration sources (the calibration block), followed by 5 s of idle for telescope stability before the science observation is executed. As the telescope is scanned across the sky during science observations, all of the bolometer pixels are read out at a frequency of 40 Hz, even during periods where the telescope was turning around for the next scan leg. However, due to satellite data-rate limitations, all PACS data are averaged over four frames—effectively downsampling the data to 10 Hz. The result is a data timeline of the flux seen by each detector pixel as a function of time (and by extension, position on the sky) as the telescope is scanned over the target field.

In order to accurately reconstruct the image, two scan map AORs at orthogonal angles are required. This is because, as the telescope scans a field, the offsets of each bolometer subarray, and even each pixel, may be different from their neighbors, resulting in stripes or gradients in the final reconstructed map. However, if the same field is scanned in two orthogonal directions, many of these map artifacts can be successfully removed, by virtue of multiple different bolometers sampling each patch of the sky. Furthermore, in order to maximally sample a given sky pixel by as many bolometer pixels as possible, we chose our scan angles to be $45^\circ$ and $135^\circ$, with respect to the detector array. The orthogonal scans similarly help remove drifts in the bolometer timelines. These time-dependent variations in the detector or subarray offsets can be caused by, for example, cosmic ray hits and other instrument effects. For our survey, the typical PACS scan duration is about 200 s. However, larger maps with deeper coverage can be as long as ∼1900 s. Because two scans are needed for each target in the blue and green filter, there are two pairs of scans and cross-scans in the red channel, giving us better sensitivity. Unfortunately, due to unforeseen consequences, several galaxy components in IRAS F02071–1023 and IRAS F07256+3355 had sufficient coverage from only one of the scans, which resulted in more noise along the scan direction around the target.

Because, by definition, all of the objects in the GOALS sample have an IRAS 60 μm flux of at least 5.24 Jy, the galaxies or galaxy systems are bright enough that only one repetition was needed for each PACS scan and cross-scan. With one pair of scan and cross-scan observations, we achieved a 1-σ point source sensitivity of approximately 4 mJy in the central area, and approximately 8 mJy averaged over the entire map for both blue and green observations. By combining all four red channel scans and cross-scans, we achieved a 1-σ point source sensitivity of about 6 mJy in the central area, and about 12 mJy averaged over the entire map. On the other hand, the extended flux sensitivities for one repetition (one scan and cross-scan pair) are 5.3 MJy sr⁻¹, 5.2 MJy sr⁻¹, and 1.7 MJy sr⁻¹ for the 70 μm, 100 μm, and 160 μm channels, respectively.

SPIRE (Griffin et al. 2010) is a submillimeter camera on Herschel that operates between the 194 and 671 μm wavelength range. In the imaging mode, it can simultaneously observe in three different broad bandpasses ($\lambda /{\rm{\Delta }}\lambda \sim 3$), centered at 250, 350, and 500 μm. Similar to PACS, SPIRE images a field by scan mapping, where the instrument field of view (4' × 8') is scanned across the sky to maximize the spatial coverage. The three detector arrays use hexagonal feedhorn-coupled bolometers, with 139, 88, and 43 bolometers for the PSW (250 μm), PMW (350 μm), and PLW (500 μm) channels, respectively. The beam profiles for each wavelength have mean FWHM values of 181, 252, and 366 for the 250, 350, and 500 μm photometer arrays, and mean ellipticities of 7%, 12%, and 9% (the beam shape changes slightly as a function of off-axis angle).

There are three main observing modes available: point source photometry, field/jiggle mapping, and scan mapping. For our observing program (dsanders_OT1_1) we chose the scan-map mode at a scan rate of 30'' s⁻¹, because it gave the best data quality as well as a larger field of view for the final map than the other two mapping modes. Nominal scan angles of 424 and 1272, with respect to the detector arrays, were used to maximize sky coverage by as many detectors as possible, as well as to minimize the effect of individual bolometer drift during data processing. Like PACS, two scans are needed for data redundancy as well as cross-linking; however, the scan and cross-scan with SPIRE are observed within a single AOR. Within our program, the vast majority of our targets were observed in the small map mode (∼150 targets), whereas the rest were taken in the large map mode (∼20 targets). The typical scan durations are ∼170 s. for small maps (∼5' × 5' guaranteed map coverage area), and up to ∼2200 s for large maps. In Figure 2 panels (b) and (d), we show two example observations using SPIRE. The top panel shows a small map mode observation, whereas the bottom panel shows a large map mode observation. In both cases, the SPIRE detector is scanned over the target coordinate (shown by the red box) from the top left to the bottom right, and then from the top right to the bottom left.

Due to the extremely sensitive design of the Herschel optics and SPIRE instrument, only one repetition was observed for every target in our observing program. The SPIRE instrument has a confusion limit of 5.8, 6.3, and 6.8 mJy beam⁻¹ for the 250, 350, and 500 μm channels, which is defined as the standard deviation of the flux density in the limit of zero instrument noise (Nguyen et al. 2010). On the other hand, the instrument noise is about 9, 7.5, and 10.8 mJy beam⁻¹ at 250, 350, and 500 μm for one repetition (scan and cross-scan) at the nominal scan speed of 30'' s⁻¹. Because many of our targets have extended features, SPIRE's 1-σ sensitivities to extended flux are at the 1.4 MJy sr⁻¹, 0.8 MJy sr⁻¹, and 0.5 MJy sr⁻¹ levels for 250, 350, and 500 μm for one repetition. These flux levels are already dominated by confusion noise, and more than sufficient to detect any cold dust components in our sample.

Table 2 below lists the observing log for our data sample. Column 1 is the galaxy reference number, and column 2 is the IRAS name of the galaxy, ordered by ascending R.A. Column 3 is the common optical counterpart names to the galaxy systems. Columns 4–7 are the observation IDs for PACS imaging. Blue corresponds to a wavelength of 70 μm, whereas green corresponds to 100 μm. Each blue and green observation pair simultaneously observes the red 160 μm channel. Two orthogonal observations are made at each wavelength to reduce imaging artifacts. We note that four galaxies in our sample do not have 100 μm observations available because they were taken from other programs that did not observe them: IRAS F02401-0013 (NGC 1068), IRAS F09320+6134 (UGC 05101), IRAS F15327+2340 (Arp 220), and IRAS F21453-3511 (NGC 7130). Column 8 is the PACS observation duration for each scan and cross-scan, unless otherwise noted. We note these are not exposure times, but instead the amount of time for each scan and cross-scan. Columns 9–10 are the observation dates for each pair of PACS scan and cross-scan, unless otherwise noted. Column 11 is the Program ID of the PACS program from which the data were obtained. We list the PID corresponding to each number in Table 2's caption. The bulk of the data (∼80%) are from OT1_sanders_1, with most of the remaining data from KPGT_esturm$\_1$ and KPOT_pvanderw_1. Column 12 is the SPIRE observation ID, which includes all three 250, 350, and 500 μm observations. The scans and cross-scans for each target are combined into one observation. Column 13 is the SPIRE observation duration, which is similar to the PACS duration. Column 14 is the SPIRE observation date, and column 15 is the PID of the SPIRE program from which the data were obtained, similar to the PACS PID column.

4.1.1. Choosing a PACS Map Maker

4.1.2. PACS Map Making with JScanam

4.2.1. Choosing a SPIRE Map Maker

4.2.2. SPIRE Map Making With 2-pass Pipeline and Destriper P0

In addition to measuring the flux, we must apply an aperture correction to account for flux outside of the aperture. The PACS aperture corrections are determined from observations of bright celestial standards, and the correction factors are included in the PACS calibration files distributed from the HSA. Within HIPE, the photApertureCorrectionPointSource task performs the aperture correction, where the input is the output product from the aperture photometry task. In addition, a responsivity version must be specified; for our data, we used the most recent version (FM 7²⁰ ). Because these aperture corrections are only applicable to point sources at each wavelength, we only apply the aperture correction to point sources within our sample. To identify the point sources, we performed PSF fitting of each source in our sample, and selected the objects with FWHM consistent with the corresponding point source FWHM in each PACS band. In Table 4, the fluxes in which an aperture correction was applied is denoted by the superscript c. Typical (average) aperture correction values for the 70, 100, and 160 μm bands are 11.7%, 12.8%, and 15.5% of the uncorrected flux, respectively. The median values of the aperture correction values are less than a percent away from the averages. We do not flag aperture-corrected fluxes in Table 3, because many of the total fluxes are a combination of aperture-corrected and uncorrected fluxes.

We also experimented with applying these corrections to marginally resolved systems and systems with a point source and extended flux. However, we found that the aperture corrections artificially boosted the flux by approximately 6% on average. This is because many of our objects have varying levels of flux contribution from the point source and extended component. Furthermore, the PACS team performed a careful surface brightness comparison²¹ of PACS data with that of IRAS and Spitzer MIPS data on the same fields. By convolving, converting, and re-gridding the higher-resolution PACS 70 μm to that of IRAS 60 μm and MIPS 70 μm, and the PACS 100 μm maps to that of IRAS 100 μm, it was shown that there is no need to apply any pixel-to-pixel gain corrections to the PACS data. They also conclude that their point-source based calibration scheme is applicable in the case for extended sources. A similar conclusion is reached for the PACS red array.²² Meléndez et al. (2014) also found in their Herschel PACS observations of the Swift BAT sample that aperture corrections on extended sources were negligible (less than 3%). Therefore, we leave sources appearing extended or semi-extended in our sample unaltered by any aperture correction.

The absolute flux calibration of PACS uses models of five different late-type standard stars, with fluxes ranging between 0.6 and 15 Jy in the three photometric bands (Balog et al. 2014). In addition, ten different asteroids are also used to establish the flux calibration over the range of 0.1–300 Jy (Müller 2014). For the standard stars, the absolute flux accuracy is within 3% at 70 and 100 μm, and within 5% at 160 μm. In addition, Uranus and Neptune were also observed for validation purposes, with fluxes of up to several hundred Jy However, a 10% reduction due to nonlinearity in the detector response was observed. Taken altogether, the error in flux calibration is consistent to within 5% of the measured flux, and takes into account flat-fielding, a responsivity correction that includes the conversion of engineering units from volts to Jy pixel⁻¹, and a gain drift correction that accounts for small drifts in gain with time (PACS Observer's Manual, and references therein). Because PACS did not perform absolute measurements over the course of the mission, the fluxes are only measured relative to the zero level calculated by the mappers, which is arbitrary.

In addition to the flux calibration uncertainty, we must also take into account the error from the background subtraction, as well as the instrumental error. The error from the background subtraction is calculated in the following manner. First, using the HIPE implementation of DAOPhot, the 1-σ dispersion is calculated from all the pixels within the background annulus surrounding the target aperture. This is then multiplied by the square root of the total number of pixels within photometry aperture, under the assumption that the error in background subtraction of individual pixels are not correlated. On the other hand, the instrumental error is calculated as the quadrature sum of all error pixels within the target aperture, using the error maps produced by the mapmaker. The total flux uncertainty is then calculated as the quadrature sum of all three error components.

We note that only two of the three galaxies in IRAS F07256+3355 (NGC 2388) were observed by PACS due to the smaller field of view, whereas SPIRE observed all three. Consequently, the total fluxes in Table 3 for this system are the sum of only the two galaxies observed by both instruments. However, SPIRE photometry of the third galaxy to the west is provided in the component flux table (Table 4). The same is also true for IRAS F23488+1949, with the third galaxy to the NNW of the closer pair.

The SPIRE 2-pass pipeline (see Section 4.2.2) produces a point-source calibrated map as the main output. However, because many of our objects appear extended or marginally extended in our sample, we followed the recommendation from the NASA Herschel Science Center (NHSC) and opted to use the extended-source calibrated maps, from which we measured all of the fluxes. Both sets of maps are produced nearly identically. However, the extended-source calibrated maps have relative gain factors applied to each bolometer's signal, which accounts for the small differences between the peak and integral of each individual bolometer's beam profile. This method helps reduce residual striping in maps with extended sources, because the relative photometric gains between all of the bolometers is properly accounted for. In addition to applying the relative gains, the PMW and PLW channels are zero-point corrected by applying a constant offset based on the Planck-HFI maps (see Section 6.2.1). The overall calibration scheme for point and extended sources is described in Griffin et al. (2013).

The primary flux calibrator for SPIRE is Neptune, chosen because it has a well-understood submillimeter/FIR spectrum and is essentially a point source in the SPIRE beams. It is also bright enough that high signal-to-noise measurements can be made from it, but not so bright that it would introduce non-linearity effects from the instrument. In order to calibrate the entire instrument, special "fine scan" observations were taken such that each bolometer was scanned across Neptune in order to absolutely calibrate each bolometer. Repeated observations of Neptune also showed that there were no statistically significant changes in the detector responses over the mission. Further details on using Neptune as the primary SPIRE flux calibrator can be found in Bendo et al. (2013).

Because the vast majority of our sources have fluxes above 30 mJy, Pearson et al. (2014) recommend using either the timeline fitter or aperture photometry. Because a significant fraction of our sample contains marginally to very extended sources, as well as point sources, we opted to measure all of our SPIRE fluxes using the annularSkyAperturePhotometry task in HIPE in order to keep our measurements as uniform as possible. However, this method results in the loss of flux outside the finite-sized aperture, for which an aperture correction is needed to fully account for all the flux. In the case of point sources, we applied the aperture correction by dividing our fluxes by the encircled energy fraction (EEF) amount corresponding to the aperture radius and SPIRE channel. The EEFs can be found in the SPIRE calibration files (accessible from within HIPE), and represent the ratio of flux (energy) inside the aperture divided by the true flux of the point source. As with the PACS aperture corrections, SPIRE fluxes in which an aperture correction was applied are denoted by a superscript c in Table 4, with average corrections of 10.1%, 10.3%, and 14.8% for the 250, 350, and 500 μm channels, respectively. Similarly the median correction values are less than a percent difference from the averages.

In order to check the validity of our point source fluxes, we measured our fluxes a second time using the timeline source fitter on a subset of 65 objects that are point sources in all three SPIRE bands. The timeline fitter is the preferred method of obtaining point source fluxes on the SPIRE maps, because it works on the baseline subtracted, destriped, and deglitched Level 1 timelines of the data (which are calibrated in Jy beam⁻¹²³ ). By using a Levenberg–Marquardt algorithm to fit a two-dimensional circular or elliptical Gaussian function to the 2D timeline data, the source can be modeled and the point source flux can be calculated from the 2D fit. The advantage is that it avoids any potential artifacts arising from the map-making process, such as smearing effects from pixelization. Because it does not use the Level 2 maps, source extraction is not necessary (i.e., aperture photometry), and there are no aperture corrections needed because the 2D fit takes into account all of the flux from the point source, in principle.

When we compared the aperture photometry results to the timeline fitter results for point sources, we found that they both agree very well at 250 and 350 μm, with average aperture/timeline flux ratios of 1.030 and 0.995, respectively. However, the 500 μm channel had a slightly lower ratio of 0.93. To further check our results, we plotted the aperture/timeline flux ratio against the aperture photometry flux for all three bands, and found no statistically significant correlation in the flux ratio as a function of flux. However, we do note that, in the 500 μm case, fluxes less than approximately 150 mJy appear to have a lower aperture/timeline flux ratio, whereas fluxes above that value have an average ratio close to unity. We believe this underestimation at faint fluxes is due to confusion noise, which was also observed in the SPIRE Map Making Test Report. Furthermore, we note that the discrepancy in the 500 μm fluxes are still consistent within the typical flux errors (∼15%). As a final check we also plotted the aperture/timeline ratio against the aperture photometry radius, and we again found no statistically significant correlation. From these tests, our point source aperture photometry fluxes appear to be in good agreement with the results from the timeline fitter.

In the case of semi-extended to extended sources, aperture corrections become more complex because the flux originates not from an unresolved source, but rather is seen as surface brightness distributed within an aperture. Although an aperture correction is needed for reasons similar to the point source case, Shimizu et al. (2016) found that their extended SPIRE fluxes for their Swift BAT sample did not need aperture corrections because they were negligible. To test this, they first convolved their 160 μm PACS data to the resolution of the three SPIRE bands, and then measured the fluxes on both the convolved and unconvolved images, using the same SPIRE aperture sizes. Aperture corrections were then calculated as the ratio of the flux on the original PACS image divided by the flux obtained on the convolved image, with resulting median aperture corrections of 1.01, 0.98, and 0.98 for the 250, 350, and 500 μm channels, respectively. This makes the assumption that the 160 μm and SPIRE fluxes originate from the same material within their galaxies. We also note that their aperture sizes are similar to ours, because their galaxy sample lies in the same redshift range. Ciesla et al. (2012) also showed, by simulating the worst-case scenario, that a maximum aperture correction of 5% is needed at 500 μm. However, this was done on an (intentionally) unphysical source that has a flat constant surface brightness, with a sharp drop to zero flux at a set radius. On more realistic sources, they calculated aperture corrections of approximately ≲2%. As these corrections are very close to unity, we follow their precedent in only reporting the integrated, background subtracted flux for our extended sources.

To calculate the flux uncertainty for the SPIRE photometry, we follow a prescription similar to what we used for PACS. The first is a systematic error in the flux calibration, related to the uncertainty in the models used for Neptune, which is the primary calibrator for SPIRE. These uncertainties, which are correlated across all three SPIRE bands, are currently quoted as 4%. The other source of uncertainty is a random uncertainty related to the ability to repeat flux density measurements of Neptune, which is 1.5% for all three bands. Altogether, these two sources of uncertainty are added linearly for a total of 5.5% error in the point source flux calibration. However, in the case of extended emission calibration, there is an additional error of 1% due to the current uncertainty in the measured beam area, which is also added linearly. This error was recently improved from 4% with the release of the SPIRE calibration version 14_2. Therefore, the total uncertainty in the extended source calibration scheme amounts to 6.5% of the background subtracted flux (Bendo et al. 2013).

To calculate our total flux uncertainties, we must also include any errors incurred from measuring and subtracting the background from the measured flux, as well as the instrumental error. To estimate the uncertainty from the background subtraction, we measure the 1-σ dispersion of the flux in each pixel within the annular area used for our background measurements. This is then multiplied by the square root of the number of pixels within the photometry aperture (which can be a fractional amount) to obtain the error in background measurement. The instrumental error is calculated by summing in quadrature the pixels within the aperture on the error map generated by the pipeline. We note that this underestimates the error, because the noise is correlated between pixels. Our final SPIRE flux uncertainties are then computed as the quadrature sum of all three sources of error. In the case where the total system flux is the sum of two (or more) components, the flux uncertainty is the quadrature sum of each galaxy component's flux error.

6.2.1. SPIRE Zero-point Correction

In Figure 5, we show the distribution of fluxes from our Herschel program in each of the three PACS and SPIRE photometer bands. The histogram x-axis range and binning for each band was selected in order to meaningfully show the data. The fluxes shown here are all 1657 measured fluxes, comprising both component and total fluxes, and do not include total system fluxes that are the sum of the component fluxes. The x-axis of each panel is shown in units of log(Jy) to encompass the wide dynamic range of fluxes measured within the data.

Zoom In Zoom Out Reset image size

As expected, the fluxes are generally higher in the three PACS bands, whereas they are lower in the SPIRE bands, due to the Rayleigh–Jeans tail of the galaxy's SED. The number of measured fluxes and bin sizes are indicated for each band, as well as the minimum and maximum fluxes. The galaxies with the highest fluxes are all nearby (IRAS F02401–0013/NGC 1068, IRAS F03316–3618/NGC 1365, and IRAS F06107+7822/NGC 2146) and tend to be quite extended in the Herschel maps—with the exception of NGC 2146, which appears to be more concentrated than the other two in the PACS 70 and 100 μm channels. On the other hand, the faintest measured fluxes in the PACS bands are well within the "faint" flux regime for PACS data reduction (see Section 4.1).

One important check is to compare our new PACS 100 μm fluxes to the legacy IRAS 100 μm fluxes published in Sanders et al. (2003), because the central wavelengths of both instruments are the same. In Figure 6, we show the filter transmission curves for PACS and IRAS in blue and red, respectively. Before comparing the fluxes measured from each telescope, several constraints must be used to ensure a meaningful comparison. Importantly, we only selected objects that either appear as single galaxies in the PACS 100 μm maps, or have component galaxies close enough that they are only marginally resolved (or not at all) by PACS. We note that the IRAS 100 μm channel has a FWHM beamsize of ∼4', which is significantly larger than the PACS 100 μm beamsize of 68. Therefore, any unresolved system in PACS would certainly appear unresolved to IRAS. Second, we also applied an aperture correction for point source objects in the PACS 100 μm maps. However, we did not apply a color correction to any of our fluxes (see Section 7.3.1). The latter point would be needed to stay in accordance with how Sanders et al. (2003) measured the IRAS RBGS fluxes (see also Soifer et al. 1989), to ensure as accurate of a comparison as possible.²⁵ Importantly, these objects span the entire range of 100 μm fluxes within the GOALS sample, and represent the entire spectrum of source morphology from point source to very extended objects.

Zoom In Zoom Out Reset image size

In the upper panel of Figure 7, we plot the 100 μm PACS/IRAS flux ratio as a function of the IRAS 100 μm flux for 128 GOALS objects that satisfy our criteria (corresponding to 64% of our sample). The red line represents the unweighted average of the ratio, which is 1.012, with dashed lines representing the 1-σ scatter of 0.09. On the other hand, the median of the PACS/IRAS ratio is 1.006. Additionally, we see no variation in the flux ratio except for fluxes above ∼100 Jy, where our PACS fluxes are slightly higher. The IRAS names of these six galaxies are F03316–3618, F06107+7822, F10257–4339, F11257+5850, 13242–5713, and F23133–4251. Of these six sources, the two with the highest PACS/IRAS ratios, F03316–3618 (NGC 1365) and F06107+7822 (NGC 2146), are large galaxies with optical sizes of 112 × 62 and 60 × 34, respectively. Their fluxes could be underestimated by IRAS because they were computed assuming point source photometry. However, once we exclude these two systems, there appears to be no PACS excess left in the bright sources. Overall, there is a broad agreement in fluxes between our Herschel data and the IRAS data, to within measurement errors (∼5%–10% for PACS).

Zoom In Zoom Out Reset image size

Additionally, we also compared the PACS 70 μm fluxes to the IRAS 60 μm fluxes. However, because of the difference in wavelength, we first had to interpolate the IRAS 60 μm measurement to 70 μm. To do this, we first estimated the power-law index to the nearest whole number on the short-wavelength side of the SED bump, using the IRAS 60 μm and PACS 70 μm fluxes.²⁶ To interpolate the IRAS 60 μm flux to 70 μm, we divided the IRAS 60 μm fluxes by multiplicative factors corresponding to each power law index found in Table 2 of the Herschel technical note PICC-ME-TN-038. These factors were calculated by the PACS team to convert PACS fluxes to other key wavelengths, and vice versa, based on SED shape. We then plotted the ratio of the PACS 70 μm flux to the interpolated IRAS 70 μm flux as a function of the IRAS flux, as shown in the bottom panel of Figure 7. The average flux ratio represented by the red line is 1.001, with a 1-σ scatter of 0.04 (dashed lines), and the median ratio is 1.00. The agreement between the PACS and IRAS data in this case is exquisite, with an even tighter relation than the 100 μm comparison throughout the entire flux range.

Another comparison is to perform a similar analysis using GOALS data from the Spitzer MIPS instrument at 70 and 160 μm (J. M. Mazzarella et al. 2017, in preparation). Unfortunately, many of the images from that program suffer from saturation and other image quality issues that make it impossible to draw a meaningful comparison. As a result, we have agreed that the PACS 70 and 160 μm data will completely supersede the corresponding MIPS data.

The results here are also similar to the analysis done in the Herschel technical note SAp-PACS-MS-0718-11, where extended source fluxes were compared between PACS to Spitzer-MIPS and IRAS. Although they found an average PACS/IRAS 100 μm flux ratio of 1.32, their dispersion in the flux ratio is very similar to our results in Figure 7. We note that their analysis was done on HIPE 6, where the PACS responsivity was not well-understood, resulting in much higher flux ratios than our result.

To check the consistency of our SPIRE fluxes, we compared the measured fluxes of our SPIRE data reduced using three different SPIRE calibrations: SPIRECAL_10_1, SPIRECAL_13_1, and the latest version SPIRECAL_14_2. In Figure 8, we show six histograms of the fractional percentage change in flux between each calibration version for each of the bands. In order to facilitate as direct of a comparison as possible, we use the uncorrected fluxes computed directly by the annularSkyAperturePhotometry task in HIPE, which are not aperture- or color-corrected. The histograms show a general trend toward longer wavelengths, and a larger variance in the percent change in flux. This is again due to the long wavelength Rayleigh–Jeans tail of the galaxy's SED, where the fainter fluxes are affected more by instrument uncertainties.

Zoom In Zoom Out Reset image size

In the histogram comparing SPIRECAL_10_1 and SPIRECAL_13_1 (Figure 8, first column), the general trend is an increase in the measured flux by an unweighted average of approximately 1.45%, 0.91%, and 1.19% of the SPIRECAL_13_1 flux, for the 250, 350, and 500 μm channels, respectively. The shape of the histogram distribution is very close to Gaussian in each case. However, the 250 μm channel shows a slight positive skewness. The main updates in the calibration and data reduction pipeline are improved absolute flux calibrations of Neptune, and a better algorithm in destriping the data and removal of image artifacts.

In the second column of Figure 8, we show the histogram of measured fluxes between SPIRECAL_13_1, and the latest version SPIRECAL_14_2. The only change was an update to the absolute flux calibration of the instrument, which resulted in an even smaller change in the average flux: 0.24%, −0.19%, and 0.25% for the 250 μm, 350 μm, and 500 μm channels respectively. SPIRE maps reduced using the two previous calibration versions are available upon request.

In this section, we detail several cautionary notes on using the data presented in this paper.

7.3.1. Color Corrections

7.3.2. PACS Saturation Limits

7.3.3. Correlated Noise in PACS Data