An important issue in determining the contribution of SMGs to the star formation history on the Universe which is unresolved is the contribution to the far-IR emission of SMGs from active nuclei. The first step in studying the relative contributions of star formation and AGN activity to the total IR luminosity is to identify which SMGs contain AGN. This is not a trivial step, even with spectroscopic data, as the studies of C05, Swinbank
ous authors (e.g., Lacy et al. 2004; Ivison et al. 2004; Sajina et al. 2005; Weedman et al. 2006) have proposed methods of separating AGN from strongly star forming galaxies in the general galaxy population based on mid-IR colors and fluxes. Several authors (e.g., Egami et al. 2004; Ivison et al. 2004; Ashby et al. 2006) have suggested that such mid-IR techniques can be used to identify powerful AGN in SMGs as well. In particular, Ivison
et al. (2004) proposed using a diagram of the ratio S24μm/S8μm plotted versus the ratio
S8μm/S4.5μm to distinguish star-forming galaxies from AGN-dominated sources and pos-
sibly even distinguish redshift. The idea is based on the observation that the SEDs of two relatively nearby ULIRGs, Arp 220 and Mrk 231, which are frequently used as templates for “Starburst ULIRGs” and “AGN-dominated ULIRGs,” cause the two galaxies to follow
well-separated tracks inS24μm/S8μm–S8μm/S4.5μm color space as the SEDs are redshifted
to larger distances. Since we now have both MIPS-24μm and IRAC data for most of the
radio-detected SMGs in the C05 sample, many of which have separate UV and optical spec- tral classifications, we can test the effectiveness of Ivison et al. (2004)’s color-color diagram in exposing AGN. Since we have secure spectroscopic redshifts for our sample, we can also test the use of the color-color diagram as a redshift indicator.
We plot the color-color diagram of Ivison et al. (2004) in Figure 5.8, including only
the SMGs with 24μm observations, which means we lack information for the galaxies in
the SSA 13 field. In the diagram we have color-coded the SMG points by redshift as in previous figures, and those objects known to show AGN features in their rest-frame UV, optical, or IR spectra, and also any classified as an AGN from X-ray observations, are circled in black. Objects with intermediate starburst/AGN spectral signatures are outlined with black squares. We have also plotted the color-color tracks for Arp 220 and Mrk 231 for
redshifts corresponding to z = 0−3, marking the location corresponding to z = 1 with a
large diagonal cross, the location of z= 2 with a large plus sign, and the location ofz= 3
with a large asterisk.
At first glance, it is apparent that the diagram will not be a good way to distinguish the redshift of SMGs; galaxies of all redshifts appear in any portion of the plot and do not appear to make a clean sequence as the tracks of Arp 220 and Mrk 231 would suggest. Our findings support the suggestion of Sajina et al. (2005) that because of large variations
in 24μm flux (as the 7.7μm PAH feature enters and leaves the bandpass) high values of
Figure 5.8 The S24μm/S8μm vs. S8μm/S4.5μm color-color plot proposed by Ivison et al.
(2004) to separate powerful AGNs from strongly star-forming systems for SMGs. The
SMGs from the sample of C05 with MIPS observations are plotted as triangles for 24μm
detections and downward arrows for sources not detected at 24μm, and are color-coded
according to redshift. The dotted line represents the color-color track of Arp 220 as its SED is redshifted, while the dashed line represents the same track for Mrk 231. Points along these
tracks corresponding to the observed colors when the SEDs are redshifted to z = (1,2,3)
are represented by the large black diagonal crosses, the large black plus signs, and the large black asterisks, respectively. SMGs with AGN spectral classifications in the literature at any wavelength are outlined by black circles, and SMGs with intermediate starburst/AGN classifications are outlined by black squares.
diagram may indeed be useful to identify AGN: 5 SMGs classified as AGN do fall on the AGN track, along with several SMGs not identified as AGN. (Note that a lack of an AGN classification in this sample does not automatically indicate that it is a starburst, since for most of our sample the classifications are only available from rest-frame UV spectra, a region in which AGN spectral diagnostics are not as clearly defined as in the optical bands). One of the previously unidentified AGNs is SMM J105200.22+572420.2; both Egami et al. (2004) and Ivison et al. (2004) suggest that it should be called an AGN on the basis of its mid-IR colors. SMM J141741.81+522823.0 is picked out as an AGN in the diagram as well; Ashby et al. (2006) suggested this object to be an AGN based on its brightness and lack of PAH features in its IRS spectrum. SMM J105155.47+572312.7 is selected out as an AGN in this diagram as well, a classification also made by Egami et al. (2004) on the basis of its “warm” SED.
However, it is important to point out that not all of the SMGs with identified AGN fall on the AGN color track. At least 7 SMGs with AGN identifications fall on the starburst track, and still more fall in between the starburst and the AGN track. The interpretation of their location in the diagram is unclear for all of these sources. One possible scenario is that the SMGs on the starburst track with spectroscopically-identified AGN are galaxies with weak, non-dominant AGN and substantial populations of low-mass stars which swamp the AGN light in the near-IR. If this were the case, we might then be able to suggest that
AGN strength increases across the plot from S8μm/S4.5μm ∼ 1−5. The SMGs falling
between the starburst and AGN tracks could then be interpreted as having AGNs with power intermediate between the galaxies on the starburst track and those on the AGN track, and we could conclude that most of our SMG sample is not dominated by a powerful, obscured AGN, even if an AGN is present.
However, Sajina et al. (2005) provide a different interpretation of the S24μm/S8μm
vs. S8μm/S4.5μm color-color plot from their simulations of mid-IR color plots based on
linear combinations of a mid-IR continuum spectrum from an AGN, a PAH emission spectrum, and a stellar continuum spectrum. The results of the simulations explain the
S24μm/S8μm–S8μm/S4.5μm color-color diagram as an extinction sequence at high redshift
in which both PAH-dominated objects and continuum-dominated objects are shifted to red-
der S8μm/S4.5μm colors as the optical depth along the line of sight is increased. Under
could possibly be PAH-dominated sources with large extinction. We might then interpret the SMGs with identified AGN which lie on the starburst track as less obscured sources than the identified AGN on the AGN sequence.
Given the different interpretations of the S24μm/S8μm vs.S8μm/S4.5μm diagram and
the ambiguities caused by identified AGN which lie in the starburst region, we suggest that
more investigation of the factors determining theS24μm/S8μm andS8μm/S4.5μm colors is
needed before this particular method of separating AGN-dominated from starburst galaxies can be widely utilized. We also suggest that the near-IR may not be the best place to separate out starbusts, since the near-IR emission of galaxies is sensitive to emission from both recently-formed stars and old, evolved stars.
5.3
Properties of Radio-Selected SMGs at 70µm
The 70μm band of MIPS samples the rest-frame 24μm emission from galaxies atz∼2, and
thus will be sensitive to warm dust at these redshifts. However, the instrument is limited by practical considerations as well as by confusion limits, such that it can only detect objects
with flux greater than ∼ 1 mJy at a 3-sigma level (Frayer et al. 2006b). Detections of
z ∼2 SMGs in this band of ∼a few mJy would likely indicate a somewhat flat SED and
the presence of a strong AGN. Detections of z < 1 SMGs at 70μm at similar flux levels,
however, will not necessarily indicate a strong AGN; rather, they will be consistent with
very cold SEDs. Here, we discuss the 70μm data for the C05 sample of SMGs presented in
Chapter 3.
We mentioned in Chapter 3 that at 70μm, a much smaller fraction of the C05 SMG sam-
ple are detected than in the IRAC bands or at 24μm. The four radio-selected SMGs detected
at 70μm, SMM J105200.22+572420.2 (z= 0.689), SMM J123634.51+621241.0 (z= 1.219),
SMM J141742.04+523025.7 (z= 0.661), and SMM J141741.81+522823.0 (z= 1.150), lie in
the low-redshift tail of the radio-detected SMG redshift distribution (see Chapter 3), which is consistent with the results of Huynh et al. (2007). Even in the GOODS-N field, which
has the deepest observations of all the SMG fields at 70μm, no high-redshift sources are
detected, so it is unlikely that the variation in coverage depth between SMG fields is entirely responsible for the low detection rate.
length, we have stacked 84×84 cutouts from the 70μm mosaics centered on the radio
positions of the SMGs undetected at 70μm, in which all sources with S/N>3 have been
subtracted. We employ a process similar to that of Huynh et al. (2007): we first subtract the median of its pixels from each cutout in the stack to improve the local background
removal and then rotate each cutout by 90◦ relative to the image before it in the stack
before coadding the images in the stack. The coaddition weights each cutout image by the RMS of its pixels. For comparison, we have also constructed a coadded, stacked image of random positions within the mosaics. To effectively represent the different depths of imag- ing between fields in the random stack, we have used in the random stack the same number of random cutouts from a particular field as the number of undetected SMGs in that field.
In the stacked image of 50 SMG source positions, we find a 3.4σ source with a flux
density of 0.48 ±0.14μJy. In comparison, in the stack of random image positions, no
sources with S/N >2 are found, suggesting that the source in the stacked SMG image is
real. However, given the range of redshift of our undetected SMG sources (0.2< z < 3.5)
and the behavior of the K-correction, it is unclear if the flux density of the source should
be interpreted as typical for all of the undetected SMGs or just a particular subset. Huynh
et al. (2007) find in their stacking analysis of SMGs at 70μm that the∼3σ source found in
their stacked image is dominated by galaxies withz <2. Accordingly, we divide our stack
into separate stacks for sources withz <2 andz >2 for comparison. Unlike the findings of
Huynh et al. (2007), no sources are detected above a S/N∼2 level in either of our low or
high-z stacks, but given the non-uniform depth of the 70μm mosaics and that a significant
number of the low-z sources in the low-z stack come from fields with less deep coverage
than the GOODS-N field, it is unclear that the stack of our low-z SMGs are any different
than that of Huynh et al. (2007).