Annu. Rev. Astron. Astrophys. 2005. 43: 727-768 Copyright © 2005 by Annual Reviews. All rights reserved

3. THE COSMIC INFRARED BACKGROUND

The CIB is the infrared part of the extragalactic background, the radiation content of the Universe today produced by galaxies at all redshifts and seen as an isotropic extragalactic background radiation. Patridge & Peebles (1967) predicted that observations of such a background could give powerful constraints on the cosmological evolution.

3.1. General Observations and Direct Cosmological Implications

The detection of the infrared part of the extragalactic background (the CIB for Cosmic Infrared Background) was the major objective of the DIRBE experiment aboard COBE. In fact, the CIB was first detected at long wavelengths by using the FIRAS spectrometer: > 200 µm (Puget et al. 1996). The CIB has subsequently been detected by DIRBE at 2.4, 3.5, 100, 140, 240 µm (see Hauser & Dwek 2001 and Kashlinsky 2005 for two reviews). The extragalactic background at 2.4 and 3.5 µm is significantly larger than that predicted by the integrated galaxy counts and their extrapolation. Similarly, the extragalactic background in the optical has been finally evaluated in combining several methods by Bernstein et al. (2002) and found to be larger than the value given by the integrated fluxes of galaxies by a factor larger than 2. In the mid-infrared, the interplanetary zodiacal dust emission is so strong that only upper limits were obtained by DIRBE. The combination of number counts by ISO/ISOCAM at 15 µm (see Elbaz & Cesarsky 2003) and by Spitzer at 24 µm (e.g., Papovich et al. 2004) giving lower limits, with the observations of TeV gamma ray emission from distant AGNs (e.g., Renault et al. 2001; Dwek & Krennrich 2005), gives a good measurement of the background at these wavelengths. The full cosmic background spectrum is shown in Figure 2. Only most recent and strongly constraining measurements have been plotted for clarity.

Figure 2. The extragalactic background over three decades in frequency from the near UV to millimeter wavelengths. Only strongly constraining measurements have been reported. We show for comparison in grey an SED of M82 (Chanial, 2003) - a starburst galaxy at L = 3 × 1010 L - normalized to the peak of the CIB at 140 µm. References for data points are given in Table 1.

Table 1. Extragalactic background references for Figure 2

Wavelength (µm)	Experiment	Measurement	Reference
0.2	FOCA	Number counts & model	Armand et al. 1994
0.30, 0.56, 0.81	HST/Las Campanas Observatory	Diffuse emission	Bernstein et al. 2002, Mattila 2003
2.2 < < 4 µm	IRTS	Diffuse emission	Matsumoto et al. 2005
2.2, 3.3	DIRBE/Lick	Diffuse emission	Gorjian et al. 2000
1.25, 2.2	DIRBE/2MASS	Diffuse emission	Wright 2001
10, 15	CAT	rays	Renault et al. 2001
15	ISOCAM	Number counts	Elbaz et al. 1999
24	Spitzer/MIPS	Number counts	Papovich et al. 2004
60	IRAS	Power spectrum	Miville-Deschênes et al. 2002
100	DIRBE	Diffuse emission	Renault et al. 2001
140, 240	DIRBE/WHAM	Diffuse emission	Lagache et al. 2000
140, 240	DIRBE	Diffuse emission	Hauser et al. 1998
850	SCUBA	Number counts	Smail et al. 2002
200 < < 1200	FIRAS	Diffuse emission	Lagache et al. 2000

Figure 2 clearly shows that the optical and infrared cosmic backgrounds are well separated. The first surprising result is that the power in the infrared is comparable to the power in the optical. In contrast, we know that locally, the infrared output of galaxies is only one third of the optical output. This implies that infrared galaxies grow more luminous with increasing z faster than do optical galaxies. A second important property to note is that the slope of the long wavelength part of the CIB, I ^1.4 (Gispert et al. 2000), is much less steep than the long wavelengths spectrum of galaxies (as illustrated in Figure 2 with the M82 SED). This implies that the millimeter CIB is not due to the millimeter emission of the galaxies that account for the peak of the CIB ( 150 µm). The implications in terms of energy output have been drawn by, e.g. Gispert et al. (2000). The infrared production rate per comoving unit volume (a) evolves faster between redshift zero and 1 than the optical one and (b) has to stay roughly constant at higher redshifts up to redshift 3 at least.

3.2. The Status of Deep Surveys: Resolved Fraction of the > 10 µm CIB

Many surveys from the mid-infrared to the millimeter have aimed to resolve the CIB into discrete sources. From short to long wavelengths the significant surveys are the following:

• ISOCAM 15 µm: Three kinds of surveys have been done. The shallowest is the ELAIS survey (European Large-Area ISO Survey, Oliver et al. 2000). The deepest is the survey in the HDF-N (Aussel et al. 1999) as well as the surveys in the direction of the galaxy clusters (e.g. Metcalfe et al. 2003). Altogether, about 1000 galaxies were detected. Above a sensitivity limit of 50 µJy, they produce a 15 µm extragalactic background light of (2.4 ± 0.5) nW m^-2 sr^-1 (Elbaz et al. 2002). This accounts for about 80% of the CIB at 15 µm based on the simplest extrapolation of the counts.

• Spitzer 24  µm: Spitzer surveys are ongoing, but more than 10⁵ sources have already been detected at 24 µm. Integrating the first number counts down to 60 µJy, a lower limit to the CIB at 24 µm of (1.9 ± 0.6) nW m^-2 sr^-1 is derived (Papovich et al. 2004). This accounts for about 70% of the CIB at 24 µm, based on a simple extrapolation of the counts.

• Spitzer 70 µm: Counts have been derived down to 15 mJy. Integrating these counts corresponds 0.95 nW m^-2 sr^-1 which explains ~ 20% of the CIB at 70 µm as derived from the Lagache et al. (2004) model (Dole et al. 2004a).

• ISOPHOT 90 µm: The most relevant data comes from the ELAIS survey that covers about 12 square degrees at 90 µm. Counts have been obtained down to 95 mJy (Héraudeau et al. 2004) resolving less than 5% of the CIB as derived from DIRBE by Renault et al. (2001).

• Spitzer 160 µm: First counts are derived down to 50 mJy. The integral of these counts corresponds to 1.4 nW m^-2 sr^-1, which explain about 7% of the CIB at 160 µm (Dole et al. 2004a).

• ISOPHOT 170 µm: Two main surveys have been conducted: the FIRBACK (Lagache & Dole 2001) and the Lockman hole (Kawara et al. 2004) surveys that covering about 5 square degrees. Counts down to 135 mJy contribute to less than 5% of the CIB (Dole et al. 2001).

• SCUBA 450 µm: Deep surveys at 450 µm are very hard to conduct from the ground. A few galaxies are detected between 10 and 50 mJy in deep surveys. Number counts down to 10 mJy give a lower limit on the CIB of about (0.7 ± 0.4) nW m^-2 sr^-1 (Smail et al. 2002). This resolves about 15% of the CIB at 450 µm.

• SCUBA 850 µm: Over 500 arcmin² of blank sky has been surveyed by several groups using SCUBA. The observations range from an extremely deep survey in the area of the HDF-N to a wider-field shallower survey. Also about 40 arcmin² of lensed cluster fields have been observed in which 17 sources have been detected. Fluxes range from about 0.5 to 8 mJy. The flux density in the resolved submillimeter population down to 1 mJy is (0.3 ± 0.1) nW m^-2 sr^-1 (Smail et al. 2002). This account for 60% of the CIB at 850 µm. Note that in deep surveys, sources with S₈₅₀ 3 mJy contribute to ~ 30% of the CIB.

• MAMBO 1200 µm: Deep blank sky areas surveyed by MAMBO cover about 500 arcmin². Number counts are given by Greve et al. (2004). By integrating the counts from 2.25 to 5.75 mJy, the resolved CIB is about 0.02 nW m^-2 sr^-1, or about 10% of the total CIB at 1200 µm.

Figure 3 shows the capabilities of the different surveys to find distant LIRGs. Spitzer observations at 24 µm are the most powerful tool to find LIRGs up to z ~ 2.2; ISOCAM was limited at z ~ 1.2. Distant ULIRGs are found by deep and large surveys at 24 and 850 µm. Note that capabilities have been computed using the model of Lagache et al. (2004). This empirical model is based on only two populations of galaxies; it aims only to model the redshift evolution of the average population. It reproduces all the observations from mid-infrared to the millimeter (see Appendix). Lewis et al. (2005) showed that a more sophisticated, bivariate SED does not much change the average properties although it does significantly change the dispersion. The Lagache et al. (2004) model is thus used in this paper as a tool to discuss observations and predictions.

Figure 3. Sensitivity to the bolometric luminosity and star-formation rate, assuming star forming galaxies of various infrared and submillimeter experiments. Detections of at least 10 sources in the surveys can be expected in the areas above the curves. We assumed the scenario of a typical deep survey (when available). IRAS 60 µm (S > 1 Jy, all sky); ISOCAM 15 µm (S > 250 µJy, 2 Sq. Deg.); ISOPHOT 170 µm (S > 180 mJy, 5 Sq. Deg.); Spitzer/MIPS 24 µm (S > 80 µJy, 5 Sq. Deg.); Spitzer/MIPS 70 µm (S > 25 mJy, 5 Sq. Deg.); Spitzer/MIPS 160 µm (S > 50 mJy, 5 Sq. Deg.); SCUBA 850 µm (S > 1 mJy, 1 Sq. Deg.). This plot makes use of the Lagache et al. (2004) model (see the Appendix).

3.3. Redshift Contribution to the CIB

From Figure 2, we see that contributions from galaxies at various redshifts are needed to fill the CIB SED shape. The bulk of the CIB in energy, i.e., the peak at about 150 µm, is not resolved in individual sources but one dominant contribution at the CIB peak can be inferred from the ISOCAM deep surveys. ISOCAM galaxies with a median redshift of ~ 0.7 resolve about 80% of the CIB at 15 µm. Elbaz et al. (2002) separate the 15 µm galaxies into different classes (ULIRGs, LIRGs, Starbursts, normal galaxies and AGNs) and extrapolate the 15 µm fluxes to 140 µm using template SEDs. A total brightness of (16 ± 5) nW m^-2 sr^-1 is found, which makes up about two thirds of the CIB observed at 140 µm by COBE/DIRBE. Hence, the galaxies detected by ISOCAM are responsible for a large fraction of energy of the CIB. About one half of the 140 µm CIB is due to LIRGs and about one third to ULIRGs. However, these ISOCAM galaxies make little contribution to the CIB in the millimeter and submillimeter. There, the CIB must be dominated by galaxies at rather high redshift for which the SED peak has been shifted. The redshift contribution to the CIB is illustrated in Figure 4. We clearly see that the submillimeter/millimeter CIB contains information on the total energy output by the high-redshift galaxies (z > 2). This is supported by the redshift distribution of the SCUBA sources at 850 µm with S₈₅₀ 3 mJy that make about 30% of the CIB and have a median redshift of 2.2 (Chapman et al. 2005).

Figure 4. Cumulative contribution to the CIB of galaxies at various redshifts from 0.5 to 8, from the model of Lagache et al. (2004). Measurements of the CIB are reported with the same symbols as in Figure 2.

Figure 5 shows the fraction of resolved CIB as a function of redshift for selected wavelengths. Fifty percents of the CIB is due to galaxies at redshift below 1 at 15 and 70 µm, 1.3 at 24 and 160 µm, 2 at 350 µm, 3 at 850 µm and 3.5 at 2000 µm (see also Table 2). It is clear that from the far-infrared to the millimeter, the CIB at longer wavelengths probes sources at higher redshifts.

Figure 5. Cumulative fraction of the CIB content as a function of redshift for various wavelengths, from the model of Lagache et al. (2004).

From Section 3.2, we see that the most constraining surveys in term of resolving the CIB are those at 15, 24 and 850 µm. Moreover, the capabilities of these surveys to find high-z objects are the best among all other existing surveys (see Figure 3). These surveys probe the CIB in well-defined and distinct redshift ranges, with median redshifts of 0.7 (Liang et al. 2004), ~ 1 (Caputi et al. 2005 and L. Yan, private communication), and 2.2 (Chapman et al. 2005) at 15, 24 and 850 µm, respectively. Such well-defined redshift ranges can be understood by looking at the K-correction. The K-correction is defined as:

(1)

where L(z = 0) is the rest-frame luminosity. This correction is specific to the spectrum of the population considered at a given luminosity and redshift. Figure 6 shows the K-correction at 15, 24, 70, 160 and 850 µm. The broad plateau observed around z = 1 at 15 µm and around z = 2 at 24 µm is caused by the PAHs' features. At longer wavelengths, the slow decrease of the K-correction is caused by the shape of the starburst spectra around the peak of their emission. At 850 µm, the monotonic rise favors the detection of high-z objects.

Figure 6. K-correction at 15, 24, 70 and 160 and 850 µm for a typical LIRG with L = 2 × 1011 L (from the model of Lagache et al. 2004).