IDENTIFICATION OF A COMPLETE 160 μm FLUX-LIMITED SAMPLE OF INFRARED GALAXIES IN THE ISO LOCKMAN HOLE 1 deg² DEEP FIELDS: SOURCE PROPERTIES AND EVIDENCE FOR STRONG EVOLUTION IN THE FIR LUMINOSITY FUNCTION FOR ULIRGs

UV/optical and NIR imaging data are critical for identifying the sources responsible for the observed FIR emission, measuring redshifts, and computing luminosities. The SWIRE data release actually includes optical (g'r'i') images obtained at Kitt Peak National Observatory (KPNO) covering a large fraction of the LH field, with extracted photometry for those sources matched to the identified IRAC counterparts of MIPS 24 μm sources (see below). However, SWIRE optical coverage is uneven across our fields, a central portion of the LHEX (ROSAT) field was not targeted for follow-up and the southwest portion of the LHNW field falls beyond the edge of the IRAC survey (see Figure 1) and consequently was also not observed in optical. As a result, almost half (19/40) of our MIPS 160 μm sources lack KPNO g'r'i' photometry in the SWIRE catalog. Fortunately, the SDSS provides full coverage of the LH field, and we therefore make use of the ugriz catalog photometry from Data Release 7 as well as display color-composite images for all of the MIPS 160 μm sources. Additionally, we use 2MASS catalogs to obtain JHK_s photometry for the majority (27/40) of our MIPS 160 μm sources.

As an initial step the 160 μm sources were matched with their counterparts in the SWIRE 70 μm and IRAC+24 μm+Optical catalogs. A series of image cutouts from each of the Spitzer IRAC+MIPS bands plus optical images from KPNO were then assembled for each source in order to identify the correct optical counterpart (see Figure 4 for an example, and Figures 4.1–4.40 in the online version of the journal for the complete sample). The red, blue, and green circles overlaid on these images indicate the position of the centroid of each detection in the 160 μm, 70 μm, and bandmerged IRAC+24 μm catalogs, respectively. Each 160 μm source has one and only one counterpart in the 70 μm images, and nearly all (36/40) have only one counterpart at 24 μm. In the cases with multiple 24 μm sources, we choose the source nearest the 160 μm centroid. Table 1 lists the source identification numbers from the SWIRE catalogs, as well as coordinates of each source taken from the SWIRE IRAC+24 μm+Optical catalog, except for sources lacking coverage in two or more IRAC bands (noted in the table) where we list coordinates from the SWIRE 24 μm catalog. Following the identification of the Optical/IRAC counterpart for each of the MIPS 160 μm sources, we then match these with objects in the 2MASS and SDSS catalogs and list their source IDs. In the case of 2MASS we prefer to quote the Extended Source Catalog (XSC) when a match is available, but also make use of the PointSource Catalog (PSC). The last two columns in Table 1 list the IDs assigned to sources by Oyabu et al. (2005) using observations from ISO. Table 2 lists the fluxes reported in the SWIRE, 2MASS, and SDSS catalogs along with our own flux measurements as described above.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (40 panels)

Tarball of figure set images

We collected spectroscopic data on the optical counterparts to the ISO sources from several sources, first of which was our library of Keck spectra obtained as part of the original ISO follow-up program which had targeted expected counterparts to the ISOPHOT 170 μm sources. Both low- and moderate-resolution spectra were taken with the Echellette Spectrograph and Imager (ESI; Sheinis et al. 2000) over several observing runs: 2000 Mar 30–31 and 2001 Jan 23–24 UT for the low-resolution data; and 2001 Feb 27–28, 2002 Jan 16–17, 2002 Feb 16, and 2002 Mar 15 for the moderate-resolution data. The low-resolution spectra had previously been used for redshift identification of putative ISOPHOT 170 μm counterparts and some of these data were published in Oyabu et al. (2005). The follow-up high-resolution spectra were obtained in order to make accurate emission-line flux measurements. These are published here for the first time. In total, we have Keck spectra for 19/40 (48%) of our complete sample. An example of our high-resolution Keck spectra is shown in Figure 5. In addition to Keck, we supplemented our redshift and flux measurements with public SDSS spectra (Adelman-McCarthy et al. 2007) for an additional nine of our targets. All 28 of the Keck + SDSS spectra are published online in Figures 5.1–5.28. In addition, we list spectroscopic redshifts for two sources for which the spectra themselves are unavailable (see Table 3).

Zoom In Zoom Out Reset image size

View extended figure (8 panels)

Ten of our MIPS 160 μm sources do not have optical spectra. Most of these sources were either not properly identified in the earlier ISOPHOT 170 μm images or had several radio and K-band counterparts within the ISOPHOT 170 μm beam where the dominant counterpart that had previously been targeted with Keck turned out not to be the correct source. For these 10 sources, we have been able to determine fairly accurate photometric redshifts, as described below.

Deep 1.4 GHz radio continuum images of the LHEX and LHNW fields were obtained using the NRAO VLA in the B-configuration in 2000 February and March-April 2001 March–April as part of the AY110 and AY121 programs. The angular resolution of the data are ∼5''. The achieved sensitivity of the LHEX data is 1σ ∼ 15 μ Jy while the sensitivity for the LHNW field is a factor of two worse because of a bright (4.2 Jy) continuum source nearby. The photometry is done using the AIPS task SAD, which fits a two-dimensional Gaussian to the brightness distribution, and total integrated flux is reported for extended sources in Table 2. A more detailed discussion of the radio data is presented by Oyabu et al. (2005).

The majority of the SWIRE 160 μm detections have spectroscopic data, which were used to determine their redshifts. For the 10 objects without spectra (noted with "Phot-z" in Table 3), we calculate a photometric redshift using the photometry from SDSS ugriz bands (except for J103341.28+580221.4, see Table 3), 2MASS JHK_s bands, and IRAC 3.6 and 4.5 μm. A χ² template-fitting method (Le Phare) was used following the prescription given in Ilbert et al. (2009). This method offers an improvement in photo-z accuracy over previous methods, due primarily to improved calibration using large spectroscopic samples from VLT-VIMOS and Keck-DEIMOS. The best-fit redshift values and uncertainties are listed in Table 3.

The photometry and redshifts were used to construct SEDs (νL_ν) for each source, which are shown in Figure 6 sorted in order of decreasing luminosity. The SEDs are characteristic of what has previously been observed for infrared-selected galaxies, with the most luminous sources showing a dominant "infrared bump" presumably due to thermal dust emission and an "optical bump" due to thermal emission from stars. Although the mid-infrared sampling is relatively sparse, it is also possible to see the effects of emission from polycyclic aromatic hydrocarbons (PAHs) in the mid-infrared at λ_rest ∼ 4–12 μm, and silicate absorption at λ_rest ∼ 10 μm.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The SEDs displayed in Figure 6 also show template fits to the data. The MIPS and 8 μm points are fit to a library of SED templates by Siebenmorgen & Krugel (2007), and the best fit is shown as solid line. The dotted line in Figure 6 represents a stellar evolution model fit to the UV–NIR data which is used to estimate stellar masses (see below). To estimate each source's total IR luminosity, L_IR(8–1000 μm), we use the prescription described by Siebenmorgen & Krugel (2007). The use of this model of SED fitting to estimate IR luminosity over others, such as Chary & Elbaz (2001) or Dale & Helou (2002), is advocated by Symeonidis et al. (2008) largely due to the tendency of these models to underestimate the peak of the FIR luminosity as represented by the 160 μm flux. This tendency of Siebenmorgen & Krugel (2007) to better fit the FIR data held true for our 160 μm sample as well. Table 3 lists the computed infrared luminosity for each source.

In addition to infrared luminosity, we use each source's flux at 160 μm, and redshift (H₀ = 75 km s⁻¹ Mpc⁻¹, Ω_m = 0.3, and Ω_Λ = 0.7 to calculate luminosity distance) to calculate an LF from our sample. Figure 7 shows the infrared LF resulting from our observations of the LH in comparison with the LF in the local universe, previously determined from the IRAS Revised Bright Galaxy Sample (RBGS) all-sky survey, which has median and maximum redshifts: z = 0.008 and z = 0.09 (Sanders et al. 2003). The LF density values and uncertainties plotted in Figure 7 are listed in Table 4. We divide the data into bins of log(L_IR/L_☉) = 0.4 in size, corresponding to steps of one in absolute magnitude. Note that SWIRE detected galaxies, J103258.0+573105 and J105349.60+570708.1 at 160 μm, but we do not include them in our analysis because their calculated total infrared luminosities are log(L_IR/L_☉) = 9.55 and 8.81 which results in them falling as the lone galaxy in their respective luminosity bins. The volume at which the survey is sensitive to galaxies below this luminosity range is small, so we restrict our attention to higher luminosity sources.

Zoom In Zoom Out Reset image size

Table 4. Luminosity Function

log(LIR)	Φ	σΦ	V/Vmax	$\sigma _{V/V_{\rm max}}$	Number	Median z
(L☉)	(Mpc−3 mag−1)
9.8	3.1 × 10−3	1.8 × 10−3	0.53	0.11	3	0.044
10.2	1.4 × 10−3	5.6 × 10−4	0.64	0.09	6	0.077
10.6	7.2 × 10−4	2.3 × 10−4	0.59	0.07	10	0.114
11.0	8.3 × 10−5	3.7 × 10−5	0.68	0.11	5	0.163
11.4	8.7 × 10−6	6.1 × 10−6	0.66	0.17	2	0.251
11.8	2.4 × 10−6	1.4 × 10−6	0.65	0.13	3	0.397
12.2	1.4 × 10−6	6.2 × 10−7	0.57	0.09	5	0.511
12.6	3.6 × 10−7	1.8 × 10−7	0.58	0.10	4	0.725

Notes. The galaxy sample is placed into luminosity bins of 0.4 in the range log(L_IR/L_☉). We list the space density Φ, the uncertainty in this value, volume sampling parameter V/V_max, uncertainty in V/V_max, the number of galaxies in each bin, and the median redshift.

Download table as:  ASCIITypeset image

In comparing the LF of our LH data with the RBGS we are comparing a narrow deep survey with 40 galaxies to a wide local survey with several hundred galaxies. This means that the relative significance of each galaxy is higher for the LH, so it is important to have accurate luminosity estimates. On the other hand, the one-magnitude bins have the effect of mitigating uncertainties in luminosity, since they can include galaxies with luminosity estimates differing by as much as a factor of 2.5. This is particularly significant for the objects in the sample that have luminosity estimates using photometric redshifts, since their luminosity is less certain than the rest of the sample. The photometric redshifts and their 68% uncertainty limits are noted in Table 3, and these limits are used to estimate uncertainties in their luminosities. In addition to photometric redshift uncertainties, the errors in matching νL_ν to a model infrared luminosity become less important when the data are binned as described. The space density of galaxies within each bin is calculated using the method developed by Schmidt (1968), which accounts for the fact that in a flux-limited sample a larger range of luminosities is observable at small distances than at greater distances. He proposes using a measure of the volume that a particular flux measurement samples, given the flux limits of the survey. For example, a galaxy at redshift z = 0.5 with a flux at 160 μm of 170 mJy could have been seen at greater redshift (and hence represent a larger volume), since the flux limit of the sample is 120 mJy. This volume sampling is characterized by the V/V_max parameter, where V is the volume corresponding to the redshift actually observed and V_max is the maximum volume over which it could be observed. A mean V/V_max value of 0.5 within a luminosity bin indicates an even distribution of galaxies within the total volume sampled in that bin.

The space density of galaxies with infrared luminosity (8–1000 μm) in the range log (L_IR/L_☉) = 10–12 appears to be consistent between the RBGS and our LH sample. In particular, Sanders et al. (2003) fit a broken power law to the RBGS sample. At log (L_IR/L_☉) = 9.5–10.5 the RBGS is fit with Φ(L) ∝ L^−0.6±0.1, and at log (L_IR/L_☉) = 10.5–12.5 the power law is Φ(L) ∝ L^−2.2±0.1. The LH data for luminosities log (L_IR/L_☉) < 12, agree within their errors to these power laws. This concurrence is to be expected given the relatively low-redshifts sampled in these lower luminosity bins. At log (L_IR/L_☉)>12.0 the situation changes. The comoving space density of ultraluminous infrared galaxies (ULIRGs: log (L_IR/L_☉)>12.0) in the log (L_IR/L_☉) = 12.0–12.4 luminosity bin is ∼7× higher in the LH than in the RBGS. The median redshift of the ULIRGs in the LH in this luminosity bin is z = 0.51. In the highest luminosity bin, log (L_IR/L_☉) = 12.4–12.8, the median redshift of the four LH galaxies in this bin is z = 0.71. To compare the comoving space density with the RGBS in this bin we extrapolate the RBGS power law to log (L_IR/L_☉) = 12.6 and find that the density in the LH sample is ∼11× higher.

Our new results for the LF of the most LIRGs in the LH are consistent with strong evolution in the comoving space density of ULIRGs. If we assume pure space-density evolution of the form (1 +  z)ⁿ, our new results for the LH imply n ∼ 6 ± 1. This is similar to what was found in an earlier study of the infrared LF of ULIRGs by Kim & Sanders (1998), where the comoving space density of ULIRGs in the IRAS 1 Jy sample (mean z ∼ 0.15) was found to be ∼2× larger than the local space density of ULIRGs in the RBGS (mean z ∼ 0.05), implying n = 7.6 ± 3.2. Our new results are also consistent with a recent determination of the extragalactic 250 μm LF by Dye et al. (2010), which shows a "smooth increase" with redshift of a factor of 3.6× in the comoving space density of luminous infrared sources between z = 0 and z = 0.2, corresponding to n = 7.1.

Deeper far-infrared surveys currently underway with Spitzer and Herschel will eventually allow us to determine whether the strong evolution observed for the most luminous infrared extragalactic sources in the relatively nearby universe continues out to higher redshift. For now, we simply note that if we assume similar strong evolution, e.g., (1 + z)⁶, in the ULIRG population out to higher redshifts, our results would imply a comoving space density of ULIRGs that is ∼700× larger at z ∼ 2 compared to the value at z = 0. Is there evidence for such a large population of ULIRGs at high redshift? The answer seems to be yes. There is a population of faint submillimeter sources detected by the Submillimeter Common User Bolometer Array (SCUBA) on the James Clerk Maxwell Telescope (JCMT), which has been interpreted variously as exotic objects, or ULIRGs at high redshift (Smail et al. 1997; Hughes et al. 1998; Barger et al. 1998; Lilly et al. 1999). Lilly et al. argued that these objects are indeed ULIRGs at z ∼ 2. Subsequently, Chapman et al. (2005) measured a range of spectroscopic redshifts, z = 1.7–2.8 for a sample of 73 submillimeter galaxies, and suggested an evolution in number density of three orders of magnitude for ULIRGs between z = 0 and z ∼ 2.5. Our results for ULIRGs in the LH, when extrapolated out to z = 2–2.5 are then consistent with the hypothesis that the SCUBA submillimeter sources are indeed ULIRGs.

To achieve a better understanding of the processes responsible for the observed infrared emission and the nature of the galaxies in our MIPS 160 μm sample, we use our UV–NIR imaging data and optical spectra to determine galaxy morphology and masses, and spectral types, respectively.

6.2.1. Imaging: Morphology and Masses

In order to develop a picture of the morphologies, and to gain an indication of the prevalence of merging/interacting galaxies in the sample we examine their UV–NIR images. We compile color composite images of the sources from those available through the Finding Chart section of the SDSS DR7 website, and show these in luminosity order in Figure 8. The zoom on these cutouts is scaled so that each box is 100 kpc on a side. At high redshifts and thus high zoom, the image quality of the SDSS charts is low, so for sources with z>0.3 we display stacked g'r'i' images from KPNO when available. A brief description of the galaxy morphologies is presented in Table 5. Many of the higher luminosity sources with log(L_IR/L_☉)>11.5 exhibit features suggestive of interactions/mergers, such as multiple cores and/or tidal tails. At luminosities lower than log(L_IR/L_☉) < 11 the large majority of sources appear to be mostly unperturbed spirals. These trends are consistent with previous studies of local samples of LIRGs and ULIRGs (e.g., Sanders & Mirabel 1996), which have shown that strong interactions and mergers appear responsible for triggering the most luminous infrared sources.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Stellar masses for each of the MIPS 160 μm sources are listed in Table 5. The masses were computed by fitting the UV–NIR SEDs using Le Phare (Ilbert et al. 2010) and assuming a Chabrier (Chabrier 2003) initial mass function (IMF). The mass range is log (M/M_☉) ∼ 10.0–11.5 corresponding to ∼0.5–3M*. Higher mass systems are more likely to be associated with higher infrared luminosity.

6.2.2. Spectroscopy: Extinction, Abundances, and Spectral Types

Our high-resolution Keck/ESI spectra, supplemented by SDSS spectra and four low-resolution ESI spectra, allow us to measure robust spectral types for 25 of the 160 μm sources in our sample. An example of these spectra was shown in Figure 5. Spectra for the 25 sources with spectral types as well as those from three sources for which we have data but were unable to measure spectral types are available in the online version of the journal. After carefully accounting for the effects of stellar absorption and applying an extinction correction (median E(B − V) = 0.7), we classify the spectra as H ii-region-like or star forming (H); composite or star formation + an AGN (C); Seyfert (S); or LINER (L), based on the classification scheme proposed by Kewley et al. (2006) (see Figure 9). Table 3 lists the measured spectral types and extinctions for each galaxy when they are available.

Zoom In Zoom Out Reset image size

Dividing the subsample with spectral types into luminosity bins, we find that 15 of 19, (79%), of galaxies in the log(L_IR/L_☉) < 11 bin have H ii-region-like spectral type, consistent with star formation as the dominant source of excitation. This is close to the H ii-region-like fraction of nearby galaxies selected at 60 μm (∼70%; Veilleux et al. 1995; Yuan et al. 2010). The other four galaxies include one Seyfert and one LINER and two objects with mixed types that suggest a "composite" starburst–active galactic nucleus (AGN) mixture of excitation. Only six galaxies with high resolution spectroscopy (and hence derived spectral types) have log(L_IR/L_☉)>11. Three (3/6 = 50%) have H ii-region-like spectral type, one is a Seyfert, one is a Seyfert/LINER, and one is a "composite" mixture of starburst/LINER excitation. Although the fraction of galaxies with H ii-region-like spectral types decreases at higher infrared luminosity (similar to what is observed for nearby galaxies selected at 60 μm), the number statistics in this high luminosity bin are too low to draw conclusions about the fractions of different spectral types. In an attempt to provide additional information on the spectral types of our high infrared luminosity sources, we have employed a new technique developed from studies of SDSS galaxies (Smolčić et al. 2008) that maps UV/optical continuum colors onto the spectral line diagnostic diagram. This method is described in the Appendix, where the SDSS photometry for all of our MIPS 160 μm sources is used to derive "P1,P2" photometric spectral types for each source, following the prescription given by Smolčić et al. (2008). These results both confirm the large H ii-region-like fraction among the lower luminosity infrared sources and show that composite and AGN spectral types appear to increase among the highest luminosity sources.

Because our spectra also contain the [O ii] λλ3727, 3729 doublet, we are able to estimate gas-phase oxygen abundances for these systems. Where available, these are listed in Table 5, using the robust [N ii]/[O ii] diagnostic of Kewley et al. (2002). To put these in context, we also used the measured K_s data to compare to the luminosity–metallicity relation in the NIR (Salzer et al. 2005). For the eight systems that have sufficient information (upper-branch R₂₃ gas abundances and measured luminosities; see Rupke et al. 2008 for more on the methodology), we find that five follow the L–Z relation of normal galaxies. Three others have higher luminosities than the data threshold and appear to be slightly below the L–Z relation (by 0.1–0.2 dex), as found for other infrared-selected objects at high luminosity (Rupke et al. 2008).

Table 5. Spectral Properties, Masses, and Morphologies

No.	Source	log(LIR)	E(B − V)	12+log[O/H]	S-Type	q-value	log(Ma)	Morphology	Notes
		(L☉)					(M☉)
35	SWIRE3_J105308.32 + 570645.6	12.74+0.19−0.44	...	...	...	>2.84	10.90	Highly disturbed	Tidal features
13	SWIRE2_24_J103538.8 + 573546	12.66+0.05−0.04	...	...	...	2.07	a	Spheroid	S overlapping foreground galaxy with z = 0.103.
8	SWIRE3_J103341.28 + 580221.4	12.62+0.10−0.15	...	...	...	2.35	a	Spheroid	SW foreground spiral with z = 0.075.
5	SWIRE2_24_J103314.8 + 573110	12.48+0.26−0.27	...	...	...	>2.66	11.25	Spheroid
1	SWIRE3_J103237.44 + 580845.9	12.28 ± 0.05	...	...	...	...	11.18	Disk	Disturbed, tidal features
12	SWIRE3_J103526.79 + 575147.6	12.25	...	...	H/C/L:	2.33	10.82	Spheroid	Pair (d = 20 kpc, Δz = 0.001); star to S.
16	SWIRE3_J103603.97 + 574812.5	12.20	...	...	...	...	10.88	Merger	Overlapping pair
28	SWIRE3_J105151.64 + 570935.7	12.12	...	...	S:	2.63	a	Spheroid	Pair ? (d = 30 kpc)
10	SWIRE2_24_J103358.9 + 572952	12.04	...	...	...	1.70	10.75	Spheroid	Tidal debris. (NE object is bkg source with z = 0.837.)
15	SWIRE2_24_J103557.1 + 572234	11.92+0.04−0.84	...	...	...	...	11.77	Disk	Large (100 kpc) edge-on disk
25	SWIRE3_J105113.41 + 571425.9	11.90	...	...	S/L:	2.06	11.49	Spheroid	Compact
23	SWIRE3_J105056.60 + 571631.2	11.69	1.00	8.87	H:	2.06	10.63	Spheroid	Tidal debris + pair (d = 40 kpc, Δz = 0.001)
9	SWIRE3_J103358.73 + 574317.1	11.46	0.88	...	H	1.84	11.70	Spheroid	Faint SE companion (d = 7 kpc)
6	SWIRE3_J103320.32 + 574913.6	11.37	0.66	8.89	H	2.28	10.77	Disk	Tidal arm(s).
31	SWIRE3_J105242.40 + 572444.7	11.05+0.07−0.15	...	...	...	2.39	10.86	Disk
36	SWIRE3_J105314.84 + 574137.6	11.04+0.21−0.15	...	...	...	2.55	10.46	Disk	Interacting system with SE companion (d = 30 kpc)
32	SWIRE3_J105252.76 + 570753.7	10.99	0.86	8.93	H	2.35	11.11	Disk	Tidal feature -X possible SW companion
19	SWIRE3_J104948.86 + 573458.2	10.89	...	...	C/S/L:	2.62	11.24	Disk	Edge-on with bright nucleus
11	SWIRE3_J103515.83 + 573337.4	10.87 ± 0.12	...	...	...	>2.96	11.04	Disk	Bright star to N
21	SWIRE3_J105041.96 + 570706.7	10.76	2.08	...	H	2.53	11.19	Disk	Edge-on
29	SWIRE3_J105207.16 + 570745.5	10.73	0.11	9.08	H	2.60	10.72	Disk
37	SWIRE3_J105318.94 + 572140.7	10.70	0.53	8.95	H	2.28	10.88	Disk
2	SWIRE2_24_J103249.4 + 573707	10.67	0.58	8.56	H	2.20	10.74	Disk
34	SWIRE3_J105301.36 + 570543.1	10.64	0.47	8.64	H	2.38	10.49	Disk	Compact with bright nucleus
7	SWIRE3_J103327.90 + 574534.4	10.60	0.71	8.87	H	>2.74	11.25	Disk	Possible tidal arm
14	SWIRE3_J103539.24 + 574243.9	10.58	0.71	8.89	H	2.58	11.03	Disk	Small companion to N
27	SWIRE3_J105150.50 + 573905.7	10.56	...	...	...	2.91	10.95	Disk	Edge-on
18	SWIRE3_J103653.50 + 575442.5	10.46	0.52	8.43	H	...	10.58	Disk	Compact with possible companion to SW
40	SWIRE3_J105432.31 + 570932.4	10.44	0.74	8.77	H	2.44	10.88	Disk	Bright nucleus
33	SWIRE3_J105256.85 + 570825.6	10.40	...	...	L:	2.45	10.80	Disk	Edge-on with companion to SE
26	SWIRE3_J105143.75 + 572936.9	10.18	1.87	...	S	2.10	11.09	Disk	Edge-on with bright nucleus
3	SWIRE3_J103253.94 + 580633.0	10.17	0.55	8.18	H	...	10.26	Disk	Disturbed
30	SWIRE3_J105225.75 + 570153.6	10.16	0.14	8.48	H	2.42	10.04	Disk	Blue compact
24	SWIRE3_J105100.42 + 574114.9	10.12	1.41	...	H	>2.93	10.56	Disk	Face-on
38	SWIRE3_J105320.92 + 571433.2	10.03	...	...	...	>3.09	11.47	Disk	Face-on, disturbed with bright nucleus
20	SWIRE3_J104956.06 + 571440.4	9.98+0.37−0.46	...	...	...	>3.06	10.23	Disk	Small, edge-on
17	SWIRE3_J103606.48 + 574702.4	9.95	0.55	8.58	H	...	9.70	Disk	Small, edge-on
22	SWIRE3_J105052.41 + 573506.9	9.64	0.52	8.43	H	2.92	10.40	Disk	Small, edge-on
4	SWIRE2_24_J103258.0 + 573105	9.55	0.49	9.00	H/C:	>2.88	10.47	Disk	Edge-on
39	SWIRE3_J105349.60 + 570708.1	8.81	...	...	...	2.92	9.45	Disk	Blue compact

Notes. Sources in the LHNW and LHEX fields listed in order of decreasing L_IR. Columns give R.A. ordered counter number, source ID, infrared luminosity (8–1000 μm) in units of L_☉, extinction, metallicity, spectral type (H, star-forming; C, star forming + AGN; S, Seyfert; L, LINER), stellar mass estimates in log(M_☉) using the methods described in Ilbert et al. (2010), 'q-value' of FIR-radio correlation, morphology, and Notes (tidal debris and/or companions). The spectral types are determined by use of the methods of Kewley et al. (2006), while the extinction is calculated using the Balmer decrement, and metallicities using the [N ii]/[O ii] diagnostic of Kewley et al. (2002). ^aSources J103538.8 + 573546 and J103341.28 + 580221.4 suffer from contamination due to partially overlapping foreground galaxies, which prohibits determining accurate masses. Source J105151.64 + 570935.7 is spectroscopically classified as a Seyfert and is suspected to harbor a QSO, thus the use of Le Phare to determine a mass from the UV–NIR photometry is inappropriate.

Download table as:  ASCIITypeset images: 1 2

6.2.3. Radio–FIR Correlation

The measured 1.4 GHz radio continuum fluxes are converted to 1.4 GHz radio power L_1.4GHz assuming a spectral index of α = +0.75,²⁰ and they are plotted as a function of redshift on the left panel of Figure 10. The observed 1.4 GHz radio power range between 10²⁰ and 10^24.3 W Hz⁻¹ (see Table 3), similar to the IR-selected galaxies in the local universe studied by Yun et al. (2001b), and none of the sources has sufficient radio power to be classified as a "radio-loud" object. The most luminous infrared sources also tend to be those with the largest radio luminosities, i.e., log(L_1.4GHz) = 23.0–24.4, equivalent to the radio powers typically seen among Seyfert galaxies, and thus the presence of a low luminosity AGN cannot be ruled out based on these radio powers alone.

Zoom In Zoom Out Reset image size

The well-known correlation between the measured infrared luminosity and radio power for star-forming galaxies is often quantified using the ratio commonly referred to as "q-value"

$$\begin{equation} q={\rm log} [({\rm FIR}/3.75\times 10^{12}\, {\rm Hz})/S_{1.4 {\rm GHz}}], \end{equation} \tag{ 1 }$$

where FIR is the far-infrared flux density and S_1.4GHz is in W m⁻² Hz⁻¹ (Condon 1992; Yun et al. 2001b). We computed these q-values using the L_1.4GHz derived above and L_FIR computed from the best-fit SED models integrated between λ = 40 and 500 μm, where the wavelength range has been chosen to match the original definition of L_FIR used to compute "q." As shown in the right panel in Figure 10, the derived q-values of the LH 160 μm sources fall between 1.6 and 3.0, suggesting that most of these sources follow the same radio–FIR correlation as the local star-forming galaxy populations. Some of the low-redshift (z ≲ 0.15) sources appear to have q-values on the high end of the local population. These are also the sources with the largest angular size, and the VLA measurements are likely under-estimates as a consequence. None of the LH 160 μm sources has a q-value less than 1.6 and thus a clear evidence for a radio-loud AGN.

Given that we have SDSS ugriz photometry for the complete MIPS 160 μm sample, we derive the P1 and P2 colors for all 40 galaxies in our sample by fitting their SEDs (encompassed by the SDSS ugriz photometry) with 100,000 spectra from the Bruzual & Charlot (2003) stellar population synthesis model library. Before performing the χ² minimization fit we redshift all the model spectra to the galaxy's spectroscopic redshift. The colors are then computed from the best-fit model spectrum (see Section 4.2. in Smolčić et al. 2008 for more details about the SED fitting).

To assess the accuracy of photometrically synthesized colors for IR-selected galaxies, we have derived the (P1, P2) colors via SED fitting (in the same way as described above) for an IR-selected control sample with available (P1, P2) colors computed independently from their spectra. The control sample, limited to a redshift range of 0.04 to 0.3, contains ∼1350 galaxies drawn from the SDSS DR1 "main" spectroscopic galaxy sample matched to the IRAS Faint Source Catalog (see Obrić et al. 2006 for details about the cross-correlation of the catalogs). The spectroscopically derived (P1, P2) colors have been synthesized by Smolčić et al. (2006) by convolving the SDSS high-resolution spectra with the Strömgren filter system. They estimated that these spectroscopically derived rest-frame colors are accurate to 0.03 mag.

In Figure 11, we show the difference between the photometrically and spectroscopically derived (P1, P2) colors for our IR-selected control galaxies as a function of the photometrically derived P1 color. We use the median offset as a function of P1 to correct the photometrically derived colors, i.e., to scale these to the SDSS spectroscopic system. The distribution of the (P1, P2) color differences after the corrections have been applied are shown in Figure 12. As expected, the corrections have eliminated systematic effects. Furthermore, as the accuracy of the spectroscopically derived colors has been shown to be 0.03 mag (Smolčić et al. 2006), the accuracy of the photometrically derived (P1, P2) colors is likely better than 0.12 and 0.03, respectively. It is remarkable that the accuracy of the P2 color derived via SED fitting is comparable to that of the spectroscopically derived color.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Figure 13 we show the P1 versus P2 the distribution for our 40 galaxies in our MIPS 160 μm sample (their P1, P2 colors were corrected for systematic offsets as described in the previous section; filled dots). To assess the nature of these 40 galaxies for each one we compute the probability that it is a star-forming, composite, AGN, or absorption galaxy given its (P1, P2) rest-frame colors. The probability is computed based on the underlying distribution of our ∼1300 control (SDSS–IRAS; 0.04 < z < 0.3) galaxies (spectroscopically divided into absorption, star-forming, AGN, and composite galaxies using the standard diagnostics; Baldwin et al. 1981; Kewley et al. 2001; Kauffmann et al. 2003) in the P1–P2 plane as follows. We bin the P1–P2 plane in two dimensions. The size of the bins is taken to be about two times the photometric color uncertainty (see previous section). For each (P1, P2) bin we then calculate the probability as the ratio of the number of each (spectroscopically classified) galaxy type relative to the total number of control galaxies in that bin. Given the photometrically synthesized (P1, P2) colors for our Lockman galaxies we can then access the probability of each galaxy being a star-forming, AGN, composite, or absorption galaxy. These probability contours are shown in Figure 13, and our results are summarized in Table 6.

Zoom In Zoom Out Reset image size