Annu. Rev. Astron. Astrophys. 2000. 38: 761-814 Copyright © 2000 by Annual Reviews. All rights reserved

4.1.2. Nature of ISOCAM Sources: Luminous Star-Forming Galaxies at z ~ 0.7

At the time of writing, the properties of the ISOCAM galaxies at other wavelength bands have been studied in detail in two fields: HDF (N) (Mann et al 1997, Rowan-Robinson et al 1997, Aussel et al 1999a) and the 1415+52 field of the Canada-France Redshift Survey (CFRS; Flores et al 1999a, b). In HDF(N), Aussel et al (1999a, b) have extracted a list of 38 galaxies ( 100 µJy, 99% confidence) with optical identifications in the catalog of Barger et al (1999). Of these, 26 galaxies have known redshifts. In the (10')² CFRS 1415+52 field, Flores et al (1999b) detected 78 significant ( 3) 15 µm sources ( 250 µJy), 22 of which have spectra and redshifts. The median redshift in both fields is near 0.7, and most redshifts range between 0.4 and 1.3. The ISOCAM galaxies have the optical colors of Sbc-Scd galaxies. The majority of the galaxies are disk or interacting systems, and the remainder are irregulars and E/S0s. The one elliptical in the HDF(N) sample is associated with an AGN (z = 0.96). Typical absolute K-band magnitudes HDF(N) range between -22 and -25, similar to, but on average somewhat fainter than, an L_* galaxy [M(K) = -24]. Thus, the ISOCAM galaxies are mainly luminous disk/interacting galaxies and are definitely not part of the faint blue galaxy population responsible for the excess in faint B-band counts (Ellis 1997).

The local luminosity function at the mid-IR luminosities characteristic of the ISOCAM galaxies (~ 5 × 10¹⁰ L) is dominated by AGNs (Fang et al 1998). However, the rest-wavelength B-band spectroscopy in both the HDF(N) and CFRS fields (Flores et al 1999b, Aussel et al 1999b) and the rest-wavelength R-band observations of the HDF(S) field (Rigopoulou et al 1999b) suggest that most of the ISOCAM galaxies are dominated by star formation, with no more than ~ 1/3 of them AGNs. The majority of the galaxies in the CFRS field (15% or 70%) have e(a) optical spectra ¹⁰ characteristic of post-starburst systems (age several hundred Myrs), or active starburst galaxies with large differential intrinsic dust extinction hiding the active burst, with some starburst activity prior to a few hundred million years ago and at the same time, the latter explanation is almost certainly the correct one (Aussel et al 1999b, Rigopoulou et al 1999b). For 19 galaxies with complete radio to UV SEDs, Flores et al (1999b) identified more than half as (highly reddened) starbursts. Dusty starburst galaxies such as M82 (Kennicutt 1992), many bright ULIRGs (Liu and Kennicutt 1995), and 50% of the (U)LIRG sample studied by Wu et al (1998) have e(a) spectra in the B-band. Local e(a) galaxies have large H equivalent widths, at the same time demonstrating active current star formation and differential dust extinction. Rigopoulou et al (1999b) have recently carried out near-IR, VLT-ISAAC spectroscopy of a sample of 0.6 z 1.3 ISOCAM galaxies in the HDF(S) field. The ISOCAM HDF(S) galaxies have large (50-100 Å) H equivalent widths, which provides compelling evidence that most of them are active starbursts. The simultaneous presence of heavily dust- enshrouded present star formation and less extinct older star forming activity probably indicates several starburst episodes (Sections 3.2.1, 3.4.4).

The infrared derived star formation rates are substantially greater than those determined from the [OII] (or H) lines. On the basis of the SEDs, Flores et al find that the median 8-1000 µm luminosity of the CFRS sample is ~ 3 × 10¹¹ L (star formation rate ~ 30 to 50 M yr^-1). The HDF(N) results are similar. For HDF(s) Rigopoulou et al 2000 and Franceschini 2000 (private communication) deduced star formation rates of a few tens of M yr^-1 from the H emission but typically three times greater values from the mid-IR data. Most of the faint ISOCAM galaxies thus appear to be LIRGs. A smaller fraction (~ 25%) of the CFRS sources have ULIRG-like luminosities (10¹² L), in agreement with the work of Rowan-Robinson et al (1997) for HDF(N). Other confirmations of the identification of the faint ISOCAM galaxies as luminous starbursts come from observations of several galaxies near z ~ 1 that are lensed by foreground clusters (Lemonon et al 1998, Barvainis et al 1999). All e(a) galaxies in the z = 0.2 cluster A1689 are 15 µm ISOCAM sources (Duc et al 2000). Of the k+A and e(a) galaxies in Coma, only those with emission lines have excess mid-IR emission indicative of active star formation (Quillen et al 1999).

In the LW2 filter the situation is similar, but the relative number of AGNs and ellipticals is proportionally larger. Flores et al (1999a) identified 40% of the 15 CFRS/6.75 µm galaxies with spectra as AGNs, and 53% as active starbursts or S+A galaxies. Aussel et al (1999b) classified 4 of the 6 confidently detected 6.75 µm sources in HDF(N) as elliptical galaxies.

For the great majority of ISOCAM sources, the far-IR counterparts are not yet known. Still, template spectra and the overall background constraint can be used to draw first-order conclusions (Figure 15). If all the ISOCAM sources had ULIRG spectra as cool as Arp 220 [L(80 µm) / L(8 µm) ~ 70 ¹¹], they would significantly overproduce the far-IR/submm background and would give rise to counts at 850/175 µm, well in excess of the observations (Aussel et al 1999b, Elbaz et al 2000). Thus, on average, they must have SEDs at least as warm as LIRGs [like M82 or NGC3256: L(80 µm) / L(8 µm) ~ 5..12], in which case they still produce 30-60% of the far-IR background at 140 µm (Figure 15). A similar contribution of the ISOCAM sources to the background would be expected if they were predominantly dust-enshrouded AGNs (Figure 15). This interpretation is not supported by the HDF(N,S) and CFRS samples, however.

Figure 15. Cosmic UV to mm, extragalactic background. Open squares give the lower limits from ISOCAM 15 µm, ISOPHOT 175 µm, and SCUBA 850 µm (Blain et al 1999) sources. The optical-UV points are from Pozzetti et al (1998). The COBE FIRAS (grey shaded) and DIRBE 140/240 µm (filled circles) data are from Lagache et al (1999) (from Elbaz et al 2000). Different SEDs are shown and normalized to the 15 µm ISOCAM limit: M82 (continuous line), Arp 220 (long dashes). An Arp 220-like SED would significantly overproduce the COBE background. With an M82-like background, the ISOCAM galaxies would contribute about 30% to the COBE background. The SED of a typical Seyfert 2 galaxy (and matching the COBE background) is shown as a dash-dotted line. The hatched upper limits in the mid-IR are derived from the lack of attenuation of high-energy -rays (Hauser et al 1998).

¹⁰ 'e(a)' (Poggianti and Wu 1999) or 'S+A' (Hammer et al. 1997) galaxies have moderate EW ([OII]) line emission and at the same time strong Balmer (H etc.) absorption or (at low spectral reolution) a large 3550-3850 Å Balmer continuum break, characteristic either of A-stars or very large extinction. In the A-star model the characteristic age of the (optically visible) star formation is ~ 0.5 - 1 Gyrs. Galaxies with Balmer absorption but no emission lines are called 'E+A', or 'k+A'. Back.

¹¹ Arp 220 has an extremely cool SED. Average ULIRGs have L(80 µm) / L(8 µm) ~ 25 which would still exceed the far-IR/submm background but only marginally so, yet leaving no space for a higher z population. Back.