Searching for AGN and Ultraluminous Galaxies with Spitzer

Frank Masci

Motivation:

The unique combination of long wavelength filters and sensitivity makes Spitzer the ideal telescope for selecting potential AGN candidates to high redshift which have missed detection in the optical due to obscuration by dust. The strongest evidence for a missing AGN population is suggested by the extragalactic X-ray background, where for the hardest energies, X-rays are essentially immune to dust absorption. Existing models predict that optically obscured quasars outnumber those observed by about 3:1.

Some specific questions:

What is the "true" AGN redshift distribution? How does this relate to the redshift dependence of the comoving star formation rate, or the epoch(s) where the spheroids of galaxies were assembled?

How are they connected to galaxy formation and evolution in general? What are the feedbacks, if any?

What are the implications for the local space density of supermassive (>10⁸M_sun) black holes - namely the remnants of high redshift AGN which remain undetected in the optical?

Was the universe re-ionised primarily by an early population of quasars? If so, what are the implications on theories of structure formation for an early mass spectrum dominated by supermassive black holes?

In general, how "incomplete" are existing optical quasar surveys? How well do existing samples trace the true space density of AGN at all z?

Selection Strategy:

The IRAS mission showed that AGN can be distinguished from galaxies using their "warm" infrared colors (eg. F_60µm/F_25µm<4). This is due to a characteristic excess of short wavelength emission from heated dust close to the central engine. This applies equally well to the two "orientationally defined" AGN classes: Type 1 vs Type 2, where the latter have their central engines obscured in the optical by a dusty molecular torus.

Figure 1:
By using the "MIPS" combination alone: 70µm/24µm vs. 160µm/70µm, a similar diagnostic can be applied to select z>2 candidates whose "warm" (70µm/24µm) rest frame emission is redshifted to the observed color ~160µm/70µm. Marked points along each locus are separated by z=0.2. The Type 1 and Type 2 SEDs used to make these predictions are composites derived from "local" optically-selectedAGN samples, owing to their good multiwavelength coverage (please see me if you wish to obtain these). The fact that they are based on optically selected samples is irrelevant here. They display similar warm colors as local IRAS-selected AGN of which a majority also show evidence for heavy optical extinction. The galaxy colors are from composite SEDs derived empirically by Cong Xu from a 25µm IRAS selected sample.

Figure 2:
An IRAC bandpass (eg. 5.78µm) can also be invoked to select warm Type 1 candidates. Given the ease with which to pre-select Type 1/Type 2 AGN for spectroscopic follow-up (see Fig.1), there still exist the `extreme' Type 2 quasars which may masquerade as the ubiqitous Utraluminous IR Galaxies (ULIRGs) with typically L_IR>10¹²L_sun. By including an IRAC bandpass, these can also be isolated from less luminous galaxies by their extremely red (cool) colors (see also discussion by Jason Surace). Such a relatively short wavelength filter will imply that correspondingly shorter rest wavelengths are sampled in high z sources. Such sources may thus be heavily absorbed and remain undetected in this band.Nonetheless, an IRAC "drop-out" technique may need be applied to select candidates.

Detection Limits:

Figures 3 and 4:
Shown below are predictions for flux density as a function of redshift for a "warm Type 1 AGN" (Fig.3) and "ULIRG" (Fig.4) with fixed luminosity L_24µm = 10¹²L_sun . Predictions are given in each of the MIPS bandpasses: 160µm, 70µm, 24µm and the IRAC band 5.78µm. 5-sigma flux sensitivities (in 500 sec) are indicated by horizontal bars. The main conclusion here is that ULIRGS should be detected to redshifts of 3-4 in a closed universe in all bands. The "warmer" SEDs of Type 1s enables them to be observed beyond z~6 in the 24µm and 5.78µm bands. The color-color selection technique however (eg. Fig,1) requires that they be detected at 160um. The relatively low sensitivity in this band therefore implies that Type 1s will also be detected to z ~ 4.

Number Counts and Redshift Distributions:

Assumptions:
Integral number counts and redshift distributions are computed separetely for two source populations:

1. A hypothesised population of quasars which is the sum of two sub-populations:
Optically-selected quasars with an optically derived luminosity function and evolution to z~5, and, added to this, an obscured population scaled in such a way that the total number of quasars reproduce the extragalactic X-ray background. Some assumptions are as follows:

The shape and evolution of the luminosity function is the same for obscured and unobscured objects - namely that derived from optical studies.

The ratio of obscured to unobscured ojects (in this case three) is independent of luminosity and redshift.

Single SED composites of local "warm" Type 1 and "cooler" Type 2 AGN from the far to near-IR are assumed for the unobscured and obscured objects respectively, independent of luminosity. No evolution is assumed in either of these SEDs with z.

2. A second population consisting entirely of ULIRGS (with L_24µm=10¹²L_sun) is considered separetely. These may in reality overlap significantly with the above hypothesised AGN population. Our aim is to compare how their integrated counts alone compare with those of a pure AGN population. Assumptions are:

Galaxies with L_24µm > 10¹²L_sunare selected from the local 25µm luminosity function of Shupe et al. (1998).

_{Although very little is known on the nature of their
emission, we make the naive assumption (based on my somewhat biased views)
that a majority contain buried quasars. For simplicity,} _{we
therefore assume that they evolve like optically-selected quasars to z~5.
This appears to be consistent with recent modelling of the SCUBA counts
and the far-IR} _{background, and is strikingly
similar to the dependence of cosmic star formation rate with z.}

_{A single template SED typical of a local ULIRG is assumed
to apply to all sources at all redshifts.}

_{Figure 5:}
Integral number counts are presented for both the total number of AGN (obscured + unobscured, constrained to fit the X-ray background) and galaxies with ULIRG type luminosity for the MIPS bands. All counts are integrated to z_max=5. The counts of ULIRGs appear to be less than those of AGN at 24µm due to the large difference in their SED shapes at short wavelengths. The AGNs can be detected in greater numbers to high redshift due to their flatter (warmer) rest frame SEDs (eg. Fig.6). This simple model predicts up to 1000 AGN or separetely, 1000 ULIRG systems per square degree to the 24µm sensitivity limit. However, a reliable detection across all MIPs bands will be required for their initial selection in color-color space. Thus, they may only be detected in numbers of typically 10-30 per square degree based on the 160µm sensitivity. (The plots assume H₀=75, q_o=0.5, lambda=0).

Figure 6:
The redshift distributions for ULIRGs and all AGN are shown for counts integrated to the MIPS sensitivity limits. The ULIRG counts peak at z~2-2.5, reflecting our assumption that they follow the same evolutionary rate as optically-selected quasars, which peak in space density at z~2.2. The flatter redshift distrbution observed for AGN reflects the presence of numerous low-luminosity sources which are only detectable at the lowest redshifts. It is important to note that our selection strategy discussed above is optimised for finding candidates at z~2 where warm rest frame 70um/24um colors are redshifted to ~160um/70um. Our chances for selecting candidates are therefore maximised.

Conclusions:

1. Warm Type1/Type2 AGN candidates to z ~ 4 will be relatively easy to select using the three MIPS bands for further study. Independent of the amount of optical extinction, (cf. the warm IRAS-selected AGN), their warm colors provide the essential criterion.
2. The warmest AGN (predominately Type 1s) can also be selected by combining the MIPS bands with an IRAC band (eg. the 5.8µm band). Such a combination however is more optimised at isolating ULIRG-like objects to z~4 through their "cool" colors.
3. Simple models of the total number of AGN constrained to fit the X-ray background predict about 10 AGN per square degree in the combined MIPS bands (limited by the 160µm sensitivity). Likewise, the surface density of galaxies expected with ULIRG type luminosity is about 30 per square degree - assuming they evolve like optical quasars! If so, the simple model above predicts that up to 30% of ULIRG detections at 160µm may be associated with AGN.
4. Spitzer will be the best "warm" AGN and ULIRG detector on the market. However, a determination of the true fraction of AGN-powered ULIRGs must wait for the era of future X-ray missions like Chandra and beyond.....
Spitzer will enable a sufficiently large ULIRG sample to be compiled for such future study.

Last Updated: June 8, 1999