Flores 1999 ApJ 517 148

THE ASTROPHYSICAL JOURNAL, 517 : 148È167, 1999 May 20
( 1999. The American Astronomical Society. All rights reserved. Printed in U.S.A.
15 MICRON INFRARED SPACE OBSERV AT ORY 1 OBSERVATIONS OF THE 1415]52 CANADA-FRANCE

REDSHIFT SURVEY FIELD : THE COSMIC STAR FORMATION RATE AS DERIVED FROM DEEP
ULTRAVIOLET, OPTICAL, MID-INFRARED, AND RADIO PHOTOMETRY
H. FLORES,2 F. HAMMER,2 T. X. THUAN,3 C. CE SARSKY,4 F. X. DESERT,5 A. OMONT,6 S. J. LILLY,7 S. EALES,8
D. CRAMPTON,9 AND O. LE FÈVRE2
Received 1998 June 8 ; accepted 1998 December 28
ABSTRACT
The Canada-France Redshift Survey 1452]52 Ðeld has been deeply imaged with the Infrared Space
Observatory using ISOCAM through the LW3 Ðlter (12È18 km). Careful data analysis and comparison
with deep optical and radio data have allowed us to generate a catalog of 78 15 km sources with both
radio and optical identiÐcations. They are redder and lie at higher redshift than I-bandÈselected galaxies,
with most of them being star-forming galaxies. We have considered the galaxies detected at radio and 15
km wavelengths, which potentially include all strong and heavily extincted starbursts, up to z \ 1. Spec-
tral energy distributions (SEDs) for each of the sources have been derived using deep radio, mid-IR,
near-IR, optical, and UV photometry. The sources were then spectrally classiÐed by comparing with
SEDs of well-known nearby galaxies. By deriving their far-IR luminosities by interpolation, we can esti-
mate their star formation rate (SFR) in a way that does not depend sensitively on the extinction.
Between 35% and 85% of the star formation at z ¹ 1 is related to IR emission, and the global extinction
is in the range A \ 0.5È0.85. While heavily extincted starbursts with SFRs in excess of 100 M yr~1
V 1% of all galaxies, they contribute about 18% of the SFR density out to z \ 1._ Their
constitute less than
morphologies range from S0 to Sab, and more than a third are interacting systems. The SFR derived by
far-IR Ñuxes is likely to be D2.9 times higher than those previously estimated from UV Ñuxes. The
derived stellar mass formed since the redshift of 1 could be too high when compared with the present-
day stellar mass density. This might be due to an initial mass function in distant star-forming galaxies
di†erent from the solar neighborhood one or an underestimate of the local stellar mass density.
Subject headings : galaxies : photometry È galaxies : starburst È galaxies : stellar content È
infrared : galaxies
1. INTRODUCTION due to the extinction, which can show large variations from
one galaxy to another. For example, in IRAS star-forming
The 2800 A and [O II] 3727 emission-line luminosity galaxies, most of the energy is re-emitted at far-IR (FIR)
densities have decreased by a factor of D10 from z \ 1 to wavelengths, and these objects are either missed or their
the present day (Lilly et al. 1996 ; Hammer et al. 1997). This star formation rates are severely underestimated when
has led Madau et al. (1996) and Madau, Pozzetti, & Dick- derived by UV measurements. This is why it is often
inson (1998) to suggest that the cosmic star formation thought that the UV luminosity density is likely to provide
density has decreased by the same factor within that redshift only a lower limit to the actual star formation density. The
interval and that most of the stars seen now were formed situation is complicated further by the expected contami-
during the Ðrst half of the universeÏs existence. The UV nation by active galactic nuclei (AGNs) to the UV light
emission from galaxies is produced by a complex mix of density. In an attempt to better estimate the cosmic star
short- and moderately long-lived stars, the latter (late B and formation density, Tresse & Maddox (1998) have calculated
A0 stars) contributing more at longer UV wavelengths. the extinction-corrected Ha luminosity density at z D 0.2.
Even old stars in luminous early-type and quiescent gal- Their result is in agreement with the UV (2800 A ) at
axies can contribute to the observed UV luminosity density. z \ 0.35 (Lilly et al. 1996) if an extinction of 1 mag is
However, the most important uncertainty in estimating the assumed for the UV continuum. A preliminary study of
star formation density from the UV luminosity density is more distant galaxies indicates that the situation might be
similar at z D 1 (Glazebrook et al. 1998), but it is limited by
the difficulty of measuring the near-IR redshifted Ha line of
1 Based on observations with the Infrared Space Observatory, an ESA
project with instruments funded by ESA member states (especially the faint galaxies with 4 m telescopes. Multiwavelength
principal investigator countries : France, Germany, the Netherlands, and analyses can provide, in principle, a detailed budget of the
the United Kingdom) with the participation of the Institute for Space energy output in each wavelength range for the entire
Astronomy in Science and NASA. galaxy energy distribution. It has been shown for local gal-
2 Observatoire de Paris, Section de Meudon, DAEC, 92195 Meudon
Principal Cedex, France.
axies that FIR luminosities are tightly correlated with radio
3 Astronomy Department, University of Virginia. luminosities (Helou et al. 1988 ; Condon 1992) and bolo-
4 Service dÏAstrophysique, CEA, France. metric luminosities are most closely proportional to 12 km
5 Institut dÏAstrophysique Spatiale, Orsay. France. luminosities (Spinoglio & Malkan 1989 ; Spinoglio et al.
6 Institut dÏAstrophysique de Paris, France. 1995). These trends hold over a wide range of galaxy lumi-
7 Department of Astronomy, University of Toronto, Toronto, Canada.
8 University of Cardi†, UK. nosities, despite the large variety of galaxy energy distribu-
9 Dominion Astrophysical Observatory, National Research Council of tions. Only AGNs, which are believed to be associated to
Canada, Victoria, Canada ; Ñores=obspm.fr. supermassive black holes, appear not to follow those rela-
148
15 km ISO OBSERVATIONS 149
tions (Condon & Roderick 1988). Recent observational The relatively high fraction of AGNs is not unexpected,
advances now allow us to study distant galaxies from the because strong AGNs are generally associated with a hot
UV to the radio, sampling a wavelength range that covers dust component and have generally bluer near-IR colors
most of the domain where their energy is being emitted. than starbursts. While the 6.75 km data appears not to be
VLA deep surveys are able to detect sources down to 10 kJy optimal for selecting starbursts, they are useful for con-
(e.g., the 5 GHz surveys of Fomalont et al. 1991 ; Richards et straining galaxy spectral energy distributions (SEDs), as
al. 1998), and ISOCAM (Cesarsky et al. 1996) on board the galaxies (especially Seyfert 2 galaxies) show large variations
Infrared Space Observatory (ISO ; Kessler et al. 1996) can in their MIR color properties.
reach detection limits of 100 kJy at 15 km (Elbaz et al. We present here the 15 km LW3 observations. For gal-
1998). In the range 60È200 km, the detection limits are 0.2 axies in the redshift range 0.25 ¹ z ¹ 1, this Ðlter samples
Jy at 60km from the IRAS faint source catalog (Moshir the 8È12 km rest wavelength region. It can thus simulta-
1989) and 0.1 Jy at 175 km from the FIRBACK survey neously provide samples of normal, starburst, and active
carried out with ISO (Clements et al. 1998 ; Puget et al. galaxies that are complete to a well-deÐned bolometric Ñux
1998). FIR detections thus appear to be not sensitive limit (Spinoglio et al. 1995). MIR measurements are sensi-
enough to reach the same depth as radio and mid-IR (MIR) tive to dust thermal emission (j [ 5 km), its broad emission
deep surveys. For example, if we consider a strong and being interpreted as due to polycyclic aromatic hydrocar-
highly reddened starburst (SBH in the terminology of bons (PAHs) or unidentiÐed infrared band carriers
Schmitt et al. 1997), a S \ 250 kJy source would corre- (3 \ j \ 18 km) and nuclear nonthermal radiation. There
15km
spond to a 0.009 Jy source at 60 km, and a z \ 1 redshifted have been recent suggestions that PAHs arise from photo-
SBH with S \ 16 kJy would have 0.022 Jy at 175 km. dissociation regions around H II regions, the latter showing
5GHzand high spatial resolution of ISOCAM
The sensibility red near-IR continua (Laurent & Mirabel 1998). Genzel et
allow the study of distant Ðeld galaxies at MIR wavelengths al. (1997) have used the relative strengths of PAH features
(2 ¹ j ¹ 20 km). Star-forming galaxies and AGNs are and near-IR colors to discriminate AGN-dominated objects
easily detectable in the wavelength range 5È18 km, even at from starbursts. Lutz et al. (1998) found that the fraction of
large distances (Franceschini et al. 1991). The Canada- AGN-powered objects is relatively small at moderate IR
France Redshift Survey (CFRS) Ðeld at 1415]52 (Lilly et luminosities (typically L \ 2 ] 1012 L ) but reaches half
at higher luminosities. IR
al. 1995a) is the second most observed Ðeld at all wave- _
lengths after the Hubble Deep Field (HDF). While it does In ° 2 we discuss the data reduction, astrometry and con-
not go as deep, it is D18.5 larger in area and thus is more struction of the catalogs. In ° 3 we present the redshift
suited for source statistics when a volume-limited (z ¹ 1) distribution of the ISOCAM LW3 objects and their optical
sample is considered. It has been observed to very faint properties, including their morphologies from Hubble Space
magnitudes in the BV IK bands (photometric completeness T elescope (HST ) images. Section 4 presents SEDs from
down to I \ 23.5 mag ; Lilly et al. 1995b) and possesses UV-to-radio wavelengths and a classiÐcation scheme for
AB data for galaxies brighter than I \ 22.5
spectroscopic these and radio-selected galaxies. Derivations of the UV
mag from the CFRS and deep radio observations AB (S º and IR luminosity densities are presented in ° 5. Section 6
16 kJy ; Fomalont et al. 1991). The CFRS sample 5GHz can be describes the global star formation density at z ¹ 1 and
considered complete in the sense that it contains all lumi- compares it with previous estimates based on the UV lumi-
nous [M (AB) ¹ [20.5] galaxies in the volume out to nosity density.
z \ 1. This B paper presents a major follow-up study of the
above CFRS Ðeld, by gathering and studying representative 2. OBSERVATIONS AND DATA REDUCTION
samples of galaxies selected at radio and MIR wavelengths. The CFRS Ðeld at 1415]52 was mapped with the
With sensitivity limits of 250 kJy at 15 km (ISOCAM) and ISOCAM LW channel (PFOV 6 arcsec pixel~1) and the
16 kJy at 5 GHz (VLA), these samples should include all LW3 Ðlter (12È18 km). Twelve individual images were
strong and reddened starbursts up to z \ 1 with star forma- obtained using the microscanning astronomical obser-
tion rates larger than 100 and 70 M yr~1, respectively. vational template (AOT) mode (CAM01), resulting in a
_ total integration time of D1200 s pixel~1. The micro-
These samples should not miss any luminous FIR source in
the CFRS Ðeld, as the sources were selected using obser- scanning mode provides the best spatial resolution by
vations that cover wavelengths on either side of the 60È100 superposition of images. The same pixel of the sky was
km bump. They can thus be used ultimately to estimate the placed in di†erent parts of the camera in order to minimize
star formation density that has been missed by UV Ñux and detect any systematic e†ects. The microscanning AOT
measurements. technique also allows an accurate Ñat-Ðeld image to be gen-
The nature of the kJy radio sources in the Ðeld and their erated and yields a pixel size of 1A. 5 in the Ðnal integrated
optical counterparts has been extensively discussed by image. The detection and removal of transients and glitches,
Hammer et al. (1995). Recently, this Ðeld has been imaged integration of images, absolute Ñux calibration, and source
by ISOCAM in the LW2 (5È8.5 km) and LW3 (12È18 km) detection were carried out using the method described by
Ðlters. We have presented the 6.75 km LW2 observations in Desert et al. (1998). This method has been found to be
a previous paper (Flores et al. 1998). There we discussed the particularly well adapted to our observational strategy, i.e.,
details of the data reduction, astrometry, and conÐdence co-adding the 12 images without redundancy within each
level for each source. Fifty-four sources with a signal-to- image. Special attention was paid to possible error propa-
noise ratio (S/N) º3 were detected with S º 150 kJy, gation in the Ñux values. The photometric accuracy has
6.7km been discussed by Desert et al. (1998). From the stellar
21 of which possess spectra from the CFRS. Of the latter,
seven were stars. Among the nonstellar sources, 42% were energy distributions, we Ðnd the photometry to be accurate
classiÐed as AGNs and 50% as S]A galaxies, i.e., star- to a few percent for sources with S/N [ 10 and to D50%
forming galaxies with a signiÐcant population of A stars. for sources with S/N \ 3. Figure 1 displays the Ðnal 15 km
150 FLORES ET AL.
10
10
10
Declination (Arcmin)
Declination (Arcmin)
5
5
0
0
0
0 5 10
0 5 10 0 5 10
Right ascention (Arcmin) Right ascention (Arcmin)
FIG. 1.ÈAt left is the combined LW3 (12È18 km) image of the whole 10@ ] 10@ CFRS 1415]52 Ðeld. Center coordinates are a(2000) 14 17 53.7 and d(2000)
52 30 30.7. The image has a FWHM resolution of D11A, and the scale is 1A. 5 per pixel. In the middle is shown the map of noise. In 85% of the map the noise is
within 10% of the average. At right are shown the locations of the cataloged 15 km sources that have successfully passed our selection criteria (see text).
Sources with S/N \ 3 are represented by a single circle, while sources with S/N \ 4, 6, and 8 are shown by 2, 3, and 4 concentric circles.
image of the CFRS 1415]52 Ðeld. Individual images were Lockman Hole Deep Survey (Cesarsky et al. 1998) that
carefully registered with each other in order to optimize the show that sources with S/N º 4 in individual frames are
image quality of the brightest compact objects (see Flores et conÐrmed in 95% of cases in the Ðnal integrated image
al. 1998). The Ðnal image of the whole ISO Ðeld has a (S/N [ 10).
resolution equivalent to a median FWHMD11A (calculated Figure 2 shows the Ñux distribution, and comparison
with DAOPHOT under IRAF). It can be seen that the noise with results from other Ðelds suggests a completeness down
structure (Fig. 1) is relatively homogeneous (the standard to D350 kJy for sources with S/N º 4 and to D250 kJy for
deviation is lower than 1/10 of the mean) in D85% of the sources with S/N º 3, at least for the central 85% of the
area of the image, except near the edges. 10@ ] 10@ Ðeld. These correspond to number densities of
Point sources are iteratively extracted with a Gaussian 1590 sources deg~2, for S º 350 kJy, slightly larger
point-spread function (PSF) of 9A FWHM ; a correction than the number density of15km 1260 source deg~2 found by
factor of 1.39 (deduced from a detailed modeling of the Elbaz et al. (1998) in their ISOCAM survey of the Hubble
e†ective ISO PSF ; Cesarsky et al. 1996) is applied to the Deep Field. Surveys of low S/N ratio sources can be a†ected
measured Ñux (calibrated with ISOCAM userÏs manual by several biases, the main ones being the possible unreli-
conversion table) to account for losses in the wings of the
true PSF.
25
2.1. ISO Source Catalogs
Source detections were made on the basis of S/N and
repeatability in three independent combinations of the 12
20
individual images (for details about the source detection
repeatability and classiÐcation, see Desert et al. 1998). The
repeatability test is based on the redundancy factor, which
is the number of times that the sky pixel was seen by di†er-
15
ent pixels on the camera. The software built three indepen-
dent projection subrasters, and for each source candidate
the Ñux and error are measured at the same position in each
subrasters. The quality factor is based on Ñux measure-
10
ments and varies from the best conÐdence index (\4) to the
worst conÐdence level (\0 ; see eqs. [7]È[11] of Desert et
al. 1998). We have considered only sources with a con-
Ðdence level higher than 3, which means that the Ðnal
5
source Ñux is within 3 p (2 p in the case of 4) of the source
Ñuxes in subrasters, where p is the error in the Ðnal source
Ñux.
Altogether, 78 sources with S/N º 3 fulÐll these detection 0
criteria (Table 1). We have considered as secure detections 0 200 400 600 800 1000
those sources with S/N º 4 (41 sources ; catalogs 1, 2, and
3). The 37 sources listed in catalogs 4, 5, and 6 with
3 \ S/N \ 4 are considered ““ less ÏÏ secure. That an S/N º 4 FIG. 2.ÈFlux distribution of ISOCAM 12È18 km sources. Sources with
is a good detection criterion is conÐrmed by studies in the S/N [ 4 (catalogs 1 and 2) are shown by the shaded histogram.
TABLE 1
OPTICAL COUNTERPARTS OF ISOCAM LW3 SOURCES
ISO a d CFRS za I b V b K b dc Pd Fluxe Error

2000 2000 AB AB AB
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Catalog 1 : Objects with S/N [ 4, P \ 0.02
0 ........ 14 : 17 : 41.8 52 : 28 : 23.3 14.1157 1.150 20.54 22.45 ... 0.65 0.000000f 1653 57
5 ........ 14 : 17 : 41.9 52 : 30 : 23.2 14.1139 0.660 20.20 21.49 18.92 2.94 0.000383f 562 48
9 ........ 14 : 17 : 40.4 52 : 28 : 21.1 14.1192 ... 23.49 24.40 ... 0.92 0.012702 399 58
13 . . . . . . . 14 : 17 : 52.0 52 : 25 : 32.8 14.0855 ... 20.92 22.60 ... 0.72 0.000998 487 69
32 . . . . . . . 14 : 17 : 56.6 52 : 31 : 58.6 14.0711 ... 21.44 22.29 20.26 1.50 0.006552 274 54
42 . . . . . . . 14 : 18 : 20.9 52 : 25 : 53.0 14.0098 star 14.66 16.44 ... 1.53 0.000029 362 69
43 . . . . . . . 14 : 17 : 42.6 52 : 28 : 46.3 14.1129 ... 21.05 22.35 20.30 0.57 0.000694 209 52
44 . . . . . . . 14 : 17 : 34.9 52 : 27 : 51.0 14.1329 0.375 19.52 20.60 ... 1.89 0.000294f 347 52
51 . . . . . . . 14 : 17 : 24.3 52 : 30 : 24.0 14.1567 0.479 19.79 20.04 18.62 2.99 0.006937 459 57
84 . . . . . . . 14 : 17 : 45.8 52 : 30 : 31.2 14.1028 0.988 21.57 23.84 19.69 1.52 0.007463f 295 50
145 . . . . . . 14 : 18 : 07.1 52 : 28 : 37.3 14.0446 ... 20.00 21.33 ... 0.31 0.000089 314 54
206 . . . . . . 14 : 18 : 13.4 52 : 31 : 47.3 14.0272 0.668 20.51 21.70 19.12 1.51 0.003157 297 55
233 . . . . . . 14 : 17 : 58.5 52 : 27 : 14.8 14.0663 0.743 20.88 22.34 ... 2.51 0.002081 288 56
282 . . . . . . 14 : 18 : 18.9 52 : 29 : 05.4 14.0138 star 15.77 16.80 15.43 5.66 0.000985 214 53
294 . . . . . . 14 : 17 : 47.0 52 : 29 : 11.9 14.0998 0.430 20.58 21.84 19.06 2.44 0.008694f 245 53
303 . . . . . . 14 : 17 : 46.4 52 : 33 : 50.8 14.1006 ... 20.66 21.74 ... 2.12 0.007003 270 53
308 . . . . . . 14 : 17 : 26.7 52 : 32 : 20.9 14.1511 ... 20.71 21.61 ... 0.78 0.000990 284 61
369 . . . . . . 14 : 18 : 4.0 52 : 27 : 47.6 14.9154 0.812 21.57 23.06 ... 3.25 0.000489f 304 51
Catalog 2 : Objects with S/N [ 4, P(d, I) [ 0.02

8 ........ 14 : 17 : 23.8 52 : 27 : 49.3 14.1598 ... 19.84 21.00 ... 8.87 0.061772 367 73
24 . . . . . . . 14 : 17 : 23.7 52 : 34 : 33.7 14.1582 ... 22.78 23.44 ... 3.92 0.123123 341 77
138 . . . . . . 14 : 17 : 26.3 52 : 32 : 51.5 14.1527 ... 23.00 24.24 ... 5.54 0.268762 252 62
139 . . . . . . 14 : 18 : 18.3 52 : 29 : 16.3 14.0150 ... 22.23 24.19 19.98 1.64 0.014691 284 57
160 . . . . . . 14 : 17 : 37.4 52 : 31 : 41.9 14.1278 ... 22.08 28.36 20.31 11.04 0.448282 401 55
190 . . . . . . 14 : 17 : 54.1 52 : 33 : 56.2 14.0779 0.578 22.01 23.10 ... 2.94 0.039091 255 53
195 . . . . . . 14 : 17 : 24.6 52 : 30 : 40.1 14.1569 ... 20.61 21.65 19.51 7.05 0.071948 364 61
258 . . . . . . 14 : 17 : 41.1 52 : 30 : 22.3 14.1166 1.015 22.46 23.88 20.46 5.24 0.166123 188 47
258 . . . . . . 14.1178 9.999 22.47 24.66 20.24 6.38 0.237751 188 47
278 . . . . . . 14 : 17 : 42.9 52 : 28 : 00.8 14.1103 0.209 22.33 22.64 ... 7.84 0.306807 235 55
278 . . . . . . 14.1145 ... 22.64 25.72 ... 7.72 0.365875 235 55
278 . . . . . . 14.1091 ... 22.02 22.79 ... 10.70 0.412821 235 55
278 . . . . . . 14.1125 ... 22.62 25.75 ... 8.45 0.415531 235 55
296 . . . . . . 14 : 17 : 56.0 52 : 32 : 55.9 14.0741 ... 23.00 23.16 ... 6.09 0.314943 207 49
296 . . . . . . 14.0743 ... 21.65 22.74 ... 10.58 0.320898 207 49
304 . . . . . . 14 : 17 : 27.3 52 : 32 : 09.6 14.1489 ... 20.91 20.54 ... 9.14 0.147525 284 61
307 . . . . . . 14 : 17 : 47.9 52 : 34 : 06.6 14.0975 ... 21.21 23.19 ... 7.45 0.126166 252 55
325 . . . . . . 14 : 18 : 12.5 52 : 32 : 48.4 14.0291 ... 22.19 24.51 ... 3.88 0.077093 225 51
361 . . . . . . 14 : 17 : 49.9 52 : 33 : 43.2 14.0909 0.978 22.34 24.30 ... 2.12 0.026649 248 56
372 . . . . . . 14 : 18 : 02.7 52 : 27 : 59.6 14.0557 ... 21.63 22.51 ... 3.57 0.042434 225 52
421 . . . . . . 14 : 17 : 52.5 52 : 35 : 13.4 14.9907 ... 22.95 19.41 ... 7.78 0.447372 259 55
434 . . . . . . 14 : 17 : 44.1 52 : 25 : 50.5 14.1070 ... 20.86 21.33 ... 5.82 0.060281 277 65
440 . . . . . . 14 : 17 : 39.3 52 : 28 : 46.1 14.1232 ... 22.06 23.19 20.05 3.91 0.070782 245 55
440 . . . . . . 14.1212 ... 20.20 20.65 20.30 8.87 0.081572 245 55
457 . . . . . . 14 : 17 : 52.1 52 : 30 : 49.8 14.0846 0.989 21.81 23.15 20.34 2.97 0.034073 250 50
Catalog 3 : Objects with S/N [ 4, without optical counterpart in I

AB
021 . . . . . . 14 : 17 : 23.2 52 : 32 : 58.3 ... ... ... ... ... ... 1.0000000 270 62
202 . . . . . . 14 : 17 : 57.3 52 : 32 : 29.8 ... ... ... ... ... ... 1.0000000 209 49
288 . . . . . . 14 : 17 : 53.6 52 : 33 : 44.6 ... ... ... ... ... ... 1.0000000 231 55
439 . . . . . . 14 : 17 : 54.7 52 : 32 : 54.9 ... ... ... ... ... ... 1.0000000 248 53
Catalog 4 : Objects with 4 [ S/N [ 3, P \ 0.02
171 . . . . . . 14 : 18 : 16.1 52 : 29 : 39.2 14.0198 1.603 20.04 20.21 19.86 1.52 0.002196 211 56
183 . . . . . . 14 : 17 : 30.9 52 : 33 : 44.0 14.1400 star 15.61 16.13 ... 5.59 0.000845 208 60
228 . . . . . . 14 : 17 : 23.3 52 : 30 : 33.4 14.1609 star 18.93 20.83 18.30 2.54 0.002518 193 61
243 . . . . . . 14 : 17 : 35.9 52 : 32 : 46.6 14.1302 0.548 20.85 21.72 ... 3.19 0.018359 188 55
310 . . . . . . 14 : 18 : 15.0 52 : 31 : 22.2 14.0227 0.772 20.84 22.08 19.47 3.42 0.020906 196 53
152 FLORES ET AL. Vol. 517
TABLE 1ÈContinued
ISO a d CFRS za I b V b K b dc Pd Fluxe Error

2000 2000 AB AB AB
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
326 . . . . . . 14 : 17 : 53.5 52 : 25 : 51.5 14.0818 0.899 21.02 22.14 ... 2.74 0.010601 165 54
351g . . . . . . 14 : 17 : 40.7 52 : 33 : 57.9 14.9504 ... 23.90 99.99 ... 8.44 0.005986f 228 59
354 . . . . . . 14 : 17 : 47.8 52 : 32 : 52.8 14.0968 ... 23.39 24.88 ... 1.07 0.015834 172 54
396 . . . . . . 14 : 18 : 18.2 52 : 33 : 04.9 14.0151 ... 17.92 18.19 ... 1.93 0.000647 185 53
408 . . . . . . 14 : 17 : 40.5 52 : 27 : 14.7 14.1190 0.754 20.99 22.49 ... 0.95 0.001837f 185 56
427 . . . . . . 14 : 17 : 59.7 52 : 26 : 01.4 14.0645 ... 22.44 24.40 ... 9.72 0.000824f 247 67
449 . . . . . . 14 : 17 : 42.7 52 : 32 : 21.6 14.1117 ... 20.79 21.15 ... 3.77 0.024364 213 54
475 . . . . . . 14 : 17 : 53.9 52 : 31 : 37.0 14.0820 0.976 21.69 24.27 19.37 11.82 0.012867f 203 53
Catalog 5 : Objects with 4 [ S/N [ 3, P [ 0.02
137 ...... 14 : 17 : 23.9 52 : 25 : 54.1 14.1597 ... 23.05 25.95 ... 4.89 0.224187 284 93
174 ...... 14 : 17 : 45.3 52 : 29 : 47.6 14.1042a 0.722 21.49 23.38 19.79 3.67 0.040133 197 50
183 ...... 14 : 17 : 30.8 52 : 33 : 54.2 14.1403 ... 21.48 24.21 ... 4.96 0.071532 208 60
228 ...... 14 : 17 : 23.7 52 : 30 : 29.7 14.1591 ... 22.76 23.65 21.60 5.51 0.225442 193 61
243 ...... 14 : 17 : 36.0 52 : 32 : 55.4 14.1300 ... 23.25 24.15 ... 5.65 0.328192 188 55
276 ...... 14 : 18 : 12.7 52 : 31 : 41.8 14.0287 ... 22.29 24.19 19.98 7.56 0.281069 175 51
277 ...... 14 : 17 : 24.7 52 : 26 : 56.2 14.1554 ... 21.90 23.07 ... 5.51 0.120360 240 76
341 ...... 14 : 17 : 23.9 52 : 32 : 28.0 14.1576 ... 22.33 23.83 ... 9.93 0.444487 214 64
341 ...... 14.1596 ... 24.21 26.44 ... 4.52 0.422721 214 64
367 ...... 14 : 17 : 25.7 52 : 31 : 32.2 14.1533 ... 22.42 23.19 21.81 7.46 0.299945 177 54
377 ...... 14 : 17 : 40.8 52 : 34 : 13.0 14.1170 ... 21.83 23.92 ... 5.14 0.100134 233 60
384 ...... 14 : 17 : 42.6 52 : 29 : 54.3 14.1120 ... 22.60 23.57 20.78 6.25 0.251090 187 51
396 ...... 14 : 18 : 17.7 52 : 33 : 02.6 14.0157 ... 22.35 23.01 ... 7.03 0.258736 185 53
399 ...... 14 : 18 : 05.8 52 : 32 : 10.2 14.0477 ... 23.01 24.73 ... 4.36 0.177522 179 53
408 ...... 14 : 17 : 40.3 52 : 27 : 20.6 14.1195 ... 22.67 24.42 ... 5.85 0.235036 185 56
424 ...... 14 : 18 : 12.3 52 : 29 : 45.4 14.0302 ... 20.87 22.20 19.22 5.07 0.046449 191 50
426 ...... 14 : 18 : 19.9 52 : 30 : 47.2 14.0112 ... 23.53 24.86 20.61 6.04 0.433874 160 53
442 ...... 14 : 17 : 38.5 52 : 34 : 43.2 14.1252 ... 21.91 21.79 ... 10.02 0.347868 221 69
449 ...... 14 : 17 : 43.3 52 : 32 : 19.5 14.1100 ... 24.46 24.37 ... 2.75 0.220019 213 54
474 ...... 14 : 17 : 31.7 52 : 29 : 11.0 14.1393 ... 22.23 23.67 21.28 2.42 0.031712 160 52
Catalog 6 : Objects with 4 [ S/N [ 3, without optical counterpart in I
AB
231 ...... 14 : 17 : 33.4 52 : 34 : 21.1 ... ... ... ... ... ... 1.0000000 209 59
268 ...... 14 : 17 : 31.4 52 : 25 : 37.5 ... ... ... ... ... ... 1.0000000 224 70
348 ...... 14 : 18 : 06.3 52 : 25 : 32.0 ... ... ... ... ... ... 1.0000000 193 67
359 ...... 14 : 17 : 23.0 52 : 28 : 59.2 ... ... ... ... ... ... 1.0000000 234 73
483 ...... 14 : 17 : 29.9 52 : 29 : 21.3 ... ... ... ... ... ... 1.0000000 167 53
a The redshift for 14.1042 was given as z \ 0.7217 by Lilly et al. (1995b). Redshifts for the four stars (denoted by ““ star ÏÏ) were determined
from additional spectra. An ellipsis indicates that the redshift is unknown.
b An ellipsis indicates that the magnitude is not available.
c Distance in arcseconds between ISO source and optical or radio counterpart.
d Probability that the coincidence is by chance.
e Flux within a 9Aaperture in kJy not corrected for aperture e†ects (see text).
f Radio source with probability based on radio-source counts.
g Hammer et al. (1995) and Fomalont et al. (1992) comment that the fainter source (I D 23.9) galaxy, which is 2A away, should be the
AB
optical counterpart. Unfortunately, no spectrum of the fainter galaxy is available.
ability of sources near the survey limit and Eddington bias. tion threshold, while they actually possess a lower Ñux
Near the Ñux density limit, completeness falls o†, and pos- density than S \ 250 kJy. We have adopted a slope of 1.7
sible false sources may be introduced. The fraction of false and perform Monte Carlo simulations of 30,000 sets,
sources can be controlled through a random match control assuming a Gaussian distribution for the noise of faint
test based on the identiÐcation rate of ISOCAM sources sources. The simulations show that D15% of the sources
with optical sources (see next section) and should be small. with S/N º 3 can be sources with Ñuxes lower than 250 kJy,
We have checked whether our higher number count relative but that fraction drops to only D2% for sources with
to the HDF data could be due to the Eddington bias. Our S/N º 4. This is probably close to the truth, since after
survey is intermediate in depth between the Lockman Hole accounting for the Eddington bias (i.e., removing 15% of the
Deep Survey (from 450 to 1400 kJy) and HDF survey (from 78 sources with S/N [ 3 as well as 2% of the 41 S/N [ 4
125 to 350 kJy). Elbaz et al. (1998) quote a slope of 1.3 and sources), the slope becomes 1.7, in good agreement with
2.1 for the HDF and Lockman Hole, respectively, and we other surveys.
should expect an intermediate slope near D1.7, lower than Biases in source counts thus a†ect mainly the faintest
the value of 2.2 derived from our S º 250 and S º 350 kJy sources in our sample (those with 3 ¹ S/N \ 4). They are
counts. We have attempted to estimate the fraction of the least powerful sources and as such cannot introduce
sources that may have entered our sample erroneously signiÐcant uncertainties on global luminosity density esti-
because statistical uncertainties bring them over the detec- mates.
No. 1, 1999 15 km ISO OBSERVATIONS 153
2.2. Astrometry and Counterparts at Radio and Optical optical image. Only 3 ^ 1 of the 78 ISOCAM sources were
W avelengths found to be randomly associated with an optical counter-
The positional accuracy of ISO sources is a†ected by part with I \ 22.5 and P \ 0.02. This should be com-
AB
both the pixel size and distortion (see Flores et al. 1998). pared with the 37 I \ 22.5 and P \ 0.02 counterparts
found in Table 1. These AB experiments suggest that non-
The superposition of the ISOCAM Ðeld onto the CFRS
1415]52 optical Ðeld was done by matching the positions Gaussian e†ects cannot a†ect our probability calculations
of the six brightest sources (which were also detected at by more than a factor of 2.
6.75 km). By comparing the astrometry of the optical Table 1 lists the sources in six catalogs, the conÐdence
and ISOCAM LW3 sources, we Ðnd a median di†erence level decreasing in each successive catalog. The ISOCAM
of D3A. 7. This is reasonable given that the ISO pixel size is LW3 sources and their optical counterparts are given in
D6A. 0. columns (1) and (2). Column (3) provides the optical source
The astrometric accuracy of the CFRS 1415]52 optical redshift when available, a star indicates that the ISO source
Ðeld is 0A. 15, based on the comparison between the optical is stellar, while an ellipsis indicates that no redshift is avail-
and radio frames (Hammer et al. 1995). We Ðrst compare able. Columns (4), (5), and (6) are the I, V , and K isophotal
the 15 km image with the VLA kJy radio map (Fomalont et magnitudes in the AB system, and an ellipsis indicates that
al. 1991) and calculate the probability of a pure coincidence, no photometry is available. Column (7) gives the angular
assuming Poisson statistics : distance in arcsec between the ISOCAM LW3 source and
optical (or radio) identiÐcation, while column (8) gives the
P(d, S ) \ 1 [ e~n(S5 GHz)nd2 , (1) associated probability of coincidence. Columns (9) and (10)
5 GHz
where d is the angular distance between the ISOCAM LW3 give the Ñux density at 12È18 km and its error in kJy. The
source and the radio source in degrees and n is the inte- positions and Ñux densities of the nine nonidentiÐed ISO
grated density of radio sources with Ñux S [dn(S )\ sources are also listed in Table 1 (catalogs 3 and 6).
5GHzare thus5GHz Among the sources with S/N º 4, only eight (19%)
83520S~1.18 dS]. Ten ISOCAM LW3 sources iden-
5GHzTable 1). They all possess optical counterparts
tiÐed (see have counterparts fainter than I \ 22.5 or no optical
(Hammer et al. 1995). counterparts. These sources couldAB be at redshifts higher
Astrometry of ISOCAM LW3 sources not detected in than 1. The fraction of very faint optical counterparts
radio has been derived by comparison with the optical increases for the fainter ISOCAM LW3 sources (41% for
sources in the CFRS 1415]52 Ðeld, and sources with 3 \ S/N \ 4), which is to be expected. On the
other hand, in the most uncertain catalog (catalog 5) one
P(d, I ) \ 1 [ e~n(IAB)nd2 (2) can expect D4.6 sources out of 19 to be pure coincidental
AB
is the probability of a pure coincidence, where n(I ) are projection. In seven cases, more than one optical counter-
integrated counts derived from the CFRS. In TableAB1, six part appear to be related to an ISOCAM LW3 source. In
catalogs have been deÐned with di†erent S/N (3 ¹ S/N ¹ 4, the following analyses, we have used the ““ best ÏÏ optical
S/N [ 4) and P values (P \ 0.02, P [ 0.02, and sources identiÐcation (i.e., those with lowest coincidental
without optical counterparts within 12A). Objects that are probability), but in Table 1 we have listed all possible
also radio sources are indicated. These all have very low optical identiÐcations with a probability within a factor of 2
probabilities of accidental coincidence. of the smallest probability.
We also compare the 15 km map with that described in
Flores et al. (1998) at 6.75 km. We found 17 of the 78 15 km
3. OPTICAL PROPERTIES OF THE 15 km COUNTERPARTS
sources to also be detected at 6.75 kJy (with Ñux densities 3.1. Color and Redshift Distributions
above 150 kJy). The fraction of 15 km sources also detected The (I[K) color distribution of the 15 km optical
at 6.75 km increases with S/N (see Table 2). AB 3) is signiÐcantly redder than that of the
counterparts (Fig.
The Ðnal noise structure in ISOCAM images is not CFRS survey (top panel), with a median 0.5 mag redder
strictly Gaussian because of possible residuals glitches and than that of CFRS galaxies [S(I[K) T D 1.3].
because the 15 km positions could not have been fully cor- AB at 15 km and with
Twenty-six of the sources detected
rected for image distortions. To calibrate the probabilities I ¹ 22.5 have spectroscopy available in the CFRS data-
given in Table 1 in an empirical way, we have applied a AB Among these are four stars that are also detected at
base.
random match control test to the ISOCAM image by rotat- 6.75 km. Figure 4 shows the redshift histogram of ISO
ing it successively by 45¡, 90¡, 180¡, and 270¡ relative to the sources superposed on the CFRS redshift distribution. The
median redshift value of the ISOCAM LW3 galaxies
TABLE 2 (SzT D 0.76) is higher than that of the CFRS (SzT D 0.58)
COUNTS OF ISO CATALOGS AT 15 km but coincides with that of the S º 16 kJy radio sources
5GHz of the 15 km optical
(Hammer et al. 1995). Since 70%
S/N º 4 4 [ S/N º 3 S/N º 3
sources have I ¹ 22.5, we estimate from the redshift
AB more than 63% of the sources with
Total . . . . . . . . . . . . . . . . . . . 41 37 78 distribution that
S º 250 kJy are at z ¹ 1.
Number I ¹ 22.5 . . . . . . 33 22 55 15km
AB
With z . . . . . . . . . . . . . . . . . . . . 17 9 26
Number I [ 22.5 . . . . . . 4 10 14
3.2. Spectral ClassiÐcations
AB
No id . . . . . . . . . . . . . . . . . . . . . 4 5 9 The spectrophotometric classiÐcation of the 15 km
optical counterparts is based on optical and emission-line
LW2 detection . . . . . . . . . . . 10 7 17
radio detection . . . . . . . . . . 5 5 10a
properties using diagnostic diagrams and spectral templates
(see Hammer et al. 1995, 1997). The following spectro-
a Radio sources with SS
5GHz
º 16 kJy. photometric types are found : 15 emission-line galaxies,
250 Most (71%) of the 15 km optical counterparts are classi-

Ðed as S]A galaxies, which are galaxies with moderate
[O II] emission (W D 20 A ) and characterized by large
200 *O II+
D(3550-3850) indices (see Table 3), as deÐned in Hammer et
150
al. (1997). Large D(3550-3850) values can be caused by
either a large A star population [D(3550-3850) is correlated
100
with the Hd equivalent width, D(3550-3850)\0.2, corre-
sponding to W (Hd) \ 5È7 A for a nonextincted galaxy] or
50
very large extinctions. As we shall see, it is the A star
hypothesis that is the most plausible. This suggests that in
0
most extragalactic 15 km sources there was star formation
occurring about 0.5 Gyr prior to the observed event. The
20 importance of the S]A population is also supported by the
fact that at high redshift, galaxies with large D(3550-3850)
15 are also detected at 6.75 km and 1.44 GHz.
10 3.3. Morphologies from HST and CFHT

About 30% of the CFRS Ðeld at 1415]52 has been
5 observed by the HST . Public domain HST images of
objects in the CFRS and AutiÐb/Low Dispersion Survey
0
Spectrograph survey (Brinchmann et al. 1997) and Groth
0 2 4 Survey (Groth et al. 1994) are available. Among the 55
I ¹ 22.5 ISO sources with optical counterparts, 16 (30%)
with have been observed with the HST F814W Ðlter. HST
FIG. 3.È(I[K) color distribution of 15 km sources with I ¹ 22.5 images of optical counterparts of kJy radio sources can be
AB
(bottom panel) compared AB (top
with the CFRS galaxy color distribution Ðnd in Hammer et al. (1996).
panel). Figure 6 displays a mosaic of HST images of optical
counterparts with I ¹ 22.5 of 15 km sources. For each
AB is indicated, as well as its morpho-
source the CFRS name
three active galaxies (including the QSO CFRS 14.0198 at logical classiÐcation, by either Brinchmann et al. (1997) or
z \ 1.6, the most distant CFRS object), one quiescent us.
galaxy, one spiral galaxy, and one H II galaxy (CFRS Two sources are unresolved by the HST (one QSO and
14.1103). Figure 5 shows the spectra of all extragalactic one H II region, 14.1103). Six galaxies (37%) are found in
objects classiÐed by their spectrophotometric type. strongly interacting systems, four (25%) are E/S0 galaxies,
and four (25%) are disk-dominated galaxies. Most of the 14
resolved sources show irregularities that might be inter-
preted as pre- or postmerging events.
4. SPECTRAL ENERGY DISTRIBUTIONS AND OBJECT

CLASSIFICATIONS
4.1. SEDs of L ocal Galaxies : T oward A T emplate Database
10
Several groups have been gathering multiwavelength
observations of local galaxies (Spinoglio et al. 1995 ; Schmitt
et al. 1998). The main limitations of the data sets are the
assumed aperture corrections and possible AGN variabil-
ity. SEDs of galaxies in the same class often show large
di†erences, especially in the infrared. This is especially true
for Seyfert 2 galaxies, which show a wide range of colors in
5 the near- and far-IR (Spinoglio et al. 1995). This large dis-
persion motivated us to use mean SEDs of local galaxies to
compare with distant sources so as to avoid biases for or
against a given class of galaxies. For example, some Seyfert
2 galaxies may be misclassiÐed as starbursts, but the e†ect
on global quantities would be compensated by true star-
burst galaxies that are misclassiÐed as Seyfert 2 galaxies.
0
However, average properties can be a†ected by a single
0 0.5 1 1.5 object with extreme properties. For example, in the Schmitt
et al. (1998) sample, the average radio luminosities for ellip-
tical and spiral galaxies appears to be overestimated
FIG. 4.ÈDashed histogram shows the redshift distribution of the 22 because of the large radio powers of NGC 1316 and NGC
identiÐcations with spectra found in the CFRS database. The shaded histo-
gram shows the redshift distribution of the sources in the catalogs 1 and 2. 598, respectively. Indeed, the mean values are signiÐcantly
The unshaded histogram shows the CFRS redshift distribution after (D8 times) larger than those derived from the much larger
rescaling. and more complete sample of local ellipticals of Wrobel &
FIG. 5.ÈSpectral classiÐcation of the 21 15 km sources with available spectra according to their continuum and line properties. For each object the CFRS
name, redshift, and spectral classiÐcation are given in the upper left corner.
Heeschen (1991). Using the median rather than the mean both local samples are a priori appropriate for use as tem-
would provide a value in much closer agreement with plates for our sources. The two local samples show some-
Wrobel & Heeschen and an SED for spirals that follows the what di†erent properties in their SEDs, especially
radio-FIR correlation (Condon 1992). In the following we concerning the average IR properties of Seyfert 2 galaxies.
will use only median points for determining the SED of The Schmitt et al. galaxies are bright infrared galaxies and
local templates. include most of the starburst and AGN templates used by
Another concern is whether using local templates to Genzel et al. (1998) to compare with ultraluminous IRAS
analyze the properties of distant sources is appropriate. We galaxies. Conversely, the Spinoglio et al. sample includes
have compared the two samples of local galaxies that have galaxies varying over a 4È5 mag range of luminosities in the
been classiÐed and studied at several wavelengths. The infrared, and their average SEDs are mostly dominated by
Schmitt et al. sample has been mainly selected from the IUE fainter galaxies. Figure 7 shows the distribution of lumi-
archives and the Spinoglio et al. sample from the 12 km nosities for the VLA/ISOCAM/CFRS objects. Rest frame
faint IRAS catalog. Our 15 km sample has been selected at luminosities at 7 km have been interpolated from broad-
the wavelengths 15/(1 ] z) and 0.835/(1 ] z) km so that band observations at 0.835, 2.2, 6.75, and 15 km and are
TABLE 3
s2 VALUES FOR ALL GALAXIES OF THE 15 km AND RADIO SAMPLES AND CLASSIFICATIONS
E Sp LINER Seyfert 2 SBL SBH Hybrid
CFRS s2 Error s2 Error s2 Error s2 Error s2 Error s2 Error s2 Error Percent SBH ] Percent S Class
14.0227 . . . . . . 12.69 1.24 3.18 0.59 2.47 0.81 0.97 0.27 2.21 0.78 0.37 0.31 0.19 0.13 9]1 SB]S
14.0272 . . . . . . 12.11 1.69 4.45 0.93 2.54 0.88 1.16 0.45 1.58 0.60 0.41 0.32 0.29 0.20 8]2 SB]S
14.0276 . . . . . . 4.22 0.31 0.75 0.22 0.75 0.18 0.51 0.17 4.10 0.85 1.32 0.20 0.15 0.12 8]2 SB]S
14.0573 . . . . . . ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... Seyfert 1
14.0663 . . . . . . 8.11 0.61 1.51 0.70 1.58 0.26 0.59 0.04 2.18 0.84 0.47 0.06 0.51 0.10 9]1 SBH
14.0727 . . . . . . 11.90 0.95 4.07 1.25 2.89 0.38 1.41 0.30 0.70 0.30 0.55 0.16 0.23 0.19 8]2 SB]S
14.0779 . . . . . . 16.16 2.25 21.11 18.57 6.07 1.31 0.55 0.17 1.73 0.54 0.63 0.30 1.40 0.67 9]1 Seyfert 2 /SBH
14.0818 . . . . . . 13.20 0.89 3.0 60.15 4.10 0.54 1.17 0.22 2.27 2.17 0.48 0.32 0.23 0.11 8]2 SB]S
14.0820 . . . . . . 2.94 2.21 0.62 0.19 1.69 1.85 2.13 0.22 16.92 4.31 6.59 0.99 0.64 0.26 1]9 S
14.0846 . . . . . . 8.94 0.74 2.17 0.22 3.65 1.16 0.45 0.19 4.85 3.24 0.83 0.29 0.35 0.22 4]6 SB]S /Seyfert2
14.0854 . . . . . . 3.46 0.43 0.73 0.22 0.94 0.40 0.21 0.10 6.47 1.27 1.42 0.11 0.06 0.13 8]2 SB]S /Seyfert2
14.0909 . . . . . . 5.13 0.97 1.51 0.33 1.55 0.24 0.15 0.01 6.36 1.05 1.67 0.21 1.46 0.56 4]6 Seyfert 2
14.0937 . . . . . . 15.61 9.20 3.63 0.72 10.81 4.20 1.53 0.16 12.68 6.98 8.53 3.37 2.79 1.11 1]9 Seyfert 2
14.0998 . . . . . . 7.89 1.75 4.42 4.22 1.42 0.19 0.86 0.18 2.92 0.23 0.51 0.22 0.31 0.20 6]4 SB]S
14.1028 . . . . . . 3.13 0.43 0.86 0.31 0.66 0.16 0.25 0.12 6.23 0.41 1.47 0.15 0.54 0.24 4]6 Seyfert 2 /SB]S
14.1041 . . . . . . 23.07 6.23 33.67 72.26 6.21 1.72 1.45 0.11 1.59 0.93 7.34 3.11 1.81 1.08 8]2 Seyfert 2 /SB]S
14.1042 . . . . . . 8.29 0.75 3.01 1.18 1.43 0.16 0.45 0.13 4.01 0.25 0.95 0.18 0.38 0.15 8]2 SB]S /Seyfert2
14.1103 . . . . . . ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... H II
14.1129 . . . . . . 13.09 2.76 2.77 0.63 3.53 1.99 0.97 0.47 1.23 0.70 0.54 0.28 0.49 0.31 7]3 SB]S
14.1139 . . . . . . 8.83 0.24 2.05 0.12 1.69 0.09 0.66 0.08 1.36 0.05 0.28 0.06 0.81 0.07 9]1 SBH
14.1190 . . . . . . 8.57 0.79 1.34 0.27 1.59 0.58 0.80 0.21 2.01 0.94 0.37 0.17 0.45 0.13 9]1 SBH
14.1302 . . . . . . ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... Seyfert 1
14.1303 . . . . . . ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... QSO
14.1329 . . . . . . 18.24 1.11 8.25 0.13 3.01 0.11 3.29 0.10 3.47 0.14 2.91 0.42 4.62 0.63 9]1 SBH /Seyfert2
14.1567 . . . . . . ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... Seyfert 1
14.9025 . . . . . . ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... LINER
14.9154 . . . . . . 25.58 0.27 6.54 0.18 12.43 0.23 1.52 0.08 12.15 0.23 5.43 0.12 5.47 0.19 4]6 Seyfert 2
g
C E S Int
14.1103 14.1157 14.1329 14.0998

Int Sab Int S0/a
14.1139 14.1190 14.1129 14.1042

S0/a Int E/S0 C
14.1028 14.0846 14.1166 14.0198

Sab S? Int? Int
14.0302 14.0711 14.1554 14.1232

FIG. 6.ÈHST 5@@ ] 5@@ images obtained with the F814W Ðlter of 16 sources detected at 15 km with I \ 22.5. For each galaxy the morphological
classiÐcation is given in the upper right corner. AB
available for most objects, except those detected only at

radio wavelengths. At z º 0.5 these galaxies have lumi-
nosities comparable with those of the Schmitt et al. galaxies
and are 1.5È2.5 mag brighter than the mean magnitude of
the Spinoglio et al. galaxies. It is reasonable to think that
the discrepancies at IR wavelengths between the Spinoglio
et al. and Schmitt et al. galaxies are mainly due to the large
IR luminosity di†erences. The large dispersions displayed
by individual Schmitt et al. galaxies in each class (especially
for Seyfert 2 galaxies) ensure that this local sample is the
most appropriate one for comparison with the intrinsically
bright distant galaxies.
4.2. ClassiÐcations of the 15 km Radio Galaxies from
T heir SEDs
According to Schmitt et al. (1997), the UV-to-optical
range constitutes the best discriminator between Seyfert 2
and SBH or SBL (starburst with low extinction) galaxies.
Seyfert 2 galaxies are generally early-type spirals, with UV
colors typically bluer than those of spirals (see, e.g., Ken-
nicutt 1992). Figure 8 shows the color distribution of the
radioÈ15 km samples. In the (B[K, 2800 A [ B) diagram,
their colors are reasonably well Ðtted by Bruzual & Charlot
(1995) models, except for three objects that are either QSOs
or Seyfert 1 galaxies. They display a range of colors that are
well Ðtted by starburst and spiral templates. The same con-
clusion can be drawn from the (B [ 7 km, 2800 A [ B) FIG. 7.ÈLogarithm of the luminosities at rest, 1.4 GHz, 7 km, and 2800
diagram, with a few objects having B [ 7 km colors consis- A against redshift. Rest frame luminosities have been interpolated from the
observed luminosities at 1.4 and 5 GHz, at 0.835, 2.2, 6.75, and 15 km, and
tent with those of Seyfert 2 templates. The latter are likely to at 8350, 5500, and 4350 A , respectively. Filled dots represent SBH, open
be associated with hot dust and have redder optical-MIR dots S]SBH, triangles Seyfert 2 or LINER, and stars QSO or Seyfert 1.
colors than starbursts. In summary, most of the galaxies Vertical arrows indicate a detection limit.
describes the current mix of old and young stellar popu-

lations in a galaxy. We have not deÐned other hybrid
classes such as the Seyfert 2] starburst class, as Schmitt et
al. have found that these two populations have rather
similar radio-to-infrared properties. Twenty-seven objects
were successfully classiÐed with this technique (the averaged
weighted s is ¹1) except for seven (Table 3). Among the
latter, four are unambiguously powerful AGNs, as evi-
denced from their optical spectra (see Fig. 5 ; Hammer et al.
1995), including a QSO (14.1303) and Seyfert 1 galaxies
(14.0573, 14.1302, and 14.1567), and one is a H II region
(14.1103). The 1 p dispersion around the s2 value for each Ðt
gives an idea of the reliability of our classiÐcation scheme
(Fig. 9 below).
4.3. Final ClassiÐcation of the z ¹ 1 15 km Radio Sources
Among the 27 objects selected either at radio or 15 km,
we Ðnd seven ambiguous cases, generally in the classes
between Seyfert 2 and S]SBH (Table 3). Table 4 sum-
marizes all the information derived from their optical
spectra and radio observations (spectral index and
imagery). For three objects, emission lines or radio proper-
ties remove the ambiguity of the classiÐcation : 14.0779
shows a low ionization spectrum, and both 14.1028 and
14.1041 have negative radio spectral indexes and radio
emission extending much beyond their optical sizes. The
validity of our classiÐcation scheme is illustrated in Figure
FIG. 8.ÈColor-color diagrams for the 15 km and radio sources. The 8, which emphasizes the relevance of the (B [ 7 km, 2800
symbols have the same meaning as in Fig. 7. In the top diagram are A [ B) color-color diagram as a diagnostic diagram for
represented galaxies with K-band photometry. In the bottom diagram
galaxies without IR measurements are excluded. In the top diagram are
distinguishing starbursts from AGNs. In that diagram, star-
shown stellar tracks from Bruzual & Charlot (1995 ; top to bottom, expo- bursts lie in a well deÐned color-color region, while AGN-
nentially decreasing SFR, with q \ 4, 1, and 0.5 Gyr, respectively). The powered sources are around the starburst region. The
dash-dotted lines represent a power law, and the extinction vector is indi- object with the smallest (B [ 7 km) color is 14.9025, which
cated. In both diagrams are shown the color-color location of Schmitt et al. is classiÐed as a LINER.
(1998) galaxy templates.
Five galaxies, all within the 15 km sample, are classiÐed
as pure SBHs. They are generally radio sources and were
classiÐed by Hammer et al. (1995) as being star-forming
detected at 15 km have colors from the UV to MIR consis- S]A objects. Figure 10 displays their SEDs, on which is
tent with those of star-forming objects : starburst and spiral superposed the SBH SED from Schmitt et al. The inter-
galaxies. polated luminosity at the 60 km bump and the 60 km lumi-
SEDs for the 27 extragalactic 15 km radio sources with nosity derived from the radio-FIR correlation (S \
z ¹ 1 have been constructed using the observed Ñuxes at 60km on
125S ; Franceschini et al. 1994) agree to within 20%
visible (B , V , and I magnitudes), near-IR (K when 5Ghz Nine objects, seven of which are in the 15 km
average.
available),ABMIR
AB (6.75 km,
AB Flores et al. 1998 ; 15 km,
AB Table
sample, are classiÐed as the superposition of a starburst
1), and radio wavelengths (1.4 and 5 GHz from Fomalont et with a spiral SED (Fig. 11). For all of them but one
al. 1991) and shifted to rest wavelengths. Each SED has (14.0846) the starburst component dominates, providing
then been compared with the average SED of well-known more than 70% of the object bolometric luminosity (see
local objects (a sample of 59 galaxies ; Schmitt et al. 1997). Table 3). Only half of these objects are detected at radio
The latter include E (ellipticals), Sp (spirals), SBHs, SBLs, wavelengths (Fig. 7), so they have likely lower star forma-
Seyfert 2 galaxies, and LINERs. Schmitt et al. (1997) also tion rates than pure SBHs. It has been widely argued that
provided standard deviations of the mean, which allows us powerful starbursts such as those discovered by IRAS could
to optimize the comparison. contain an AGN that can contribute to their infrared lumi-
In order to classify the SEDs, we have performed Monte nosities (Sanders et al. 1988). In addition to the fact that
Carlo simulations, assuming independent variations of our their energy distributions and colors are typical of those of
Ñux measurements within Gaussian error bars. In each of starbursts, it is unlikely that the starbursts described here
the 5000 Monte Carlo sets, every object SED has been Ðtted are signiÐcantly contaminated by an AGN, because : Their
to template SEDs spanning all the classes deÐned by IR luminosities (L \ 2 ] 1012 L ; see Table 4) are lower
Schmitt et al, providing a s2 value weighted by the standard than those of the IR _
ultraluminous IRAS galaxies. Local gal-
deviations of the mean given by those authors. Because axies with those luminosities are mainly star-forming gal-
galaxies often show properties intermediate between one axies (Lutz et al. 1998) ; their radio spectral indices range
class and another (see Fig. 8), we also deÐned a hybrid class from 0.4 to 1 between the observed frequencies of 1.4 and 5
that consists of a linear superposition of starburst and spiral GHz. This implies a thermal-to-nonthermal energy ratio in
SED templates, with the contribution of the spiral SED the rest frame of 0È0.2, in good agreement with the star-
varying from 10% to 90% in 10% steps. This hybrid class burst population studied by Condon (1992) ; their radio
-2
-4
-6
-8
-100
-2
-4
-6
-8
-10
8 10 12 14 8 10 12 14
FIG. 9.ÈSome examples of possible ambiguous classiÐcations of the SED. Solid dots represent Ñux measurements and are superposed to a template from
Schmitt et al. Corresponding s2 values are indicated (see text).
angular sizes (less than 2A) are always in agreement with a compact system undergoing one of its very Ðrst bursts of
their optical sizes, as expected if star formation was distrib- star formation : [O II] is not detected, while [O III]
uted over the galaxy ; and Ðve of them have z ¹ 0.7 and is very prominent. This 3727galaxy shows spectral properties5007
their spectra from CFRS have [O III] 5007 and Hb emission similar to those of a primordial galaxy, according to Tresse
lines ; they all show low ionization spectra, with a promi- et al. (1993). Those authors derived a low heavy-element
nent [O II]3727 line and, for all but one, no [O III] 5007 line. abundance (¹0.05 solar) and a large e†ective temperature
Figure 12 shows the sum of the Ðve spectra, which reveals a (T [ 50,000 K, based on the [O III] /[O III] ratio).
red spectrum with [O II] 3727/Hb D 1 and [O III] 5007]/ 4363 5007
Hb ¹ 0.05. These galaxies undoubtedly have emission-line
spectra typical of H II regions, not AGNs. 5. GLOBAL UV AND IR LUMINOSITIES
Six objects, three of which are in the 15 km sample, are
classiÐed as Seyfert 2 galaxies and four, two of which are in 5.1. Global UV L uminosity
the 15 km sample, are powerful AGNs (Seyfert 1 and QSO). L values have been interpolated from L and
2800 [or (B[I) ] colors, using a grid of models 4350 with
Their SEDs are shown in Figure 13 as well as that of the (V [I)
AB AB
radio LINER 14.9025, on which are superposed Schmitt et exponentially decreasing SFR (q \ 1 Gyr ; Bruzual &
al. templates or radio-quiet QSO SEDs from Sanders et al. Charlot 1995) and a Salpeter initial mass function [IMF ;
(1989). The two remaining objects are 14.1103, an H II dN(m)/dm \ [2.35 over m \ 0.1È100 M ]. These calcu-
galaxy, and 14.0820, an elliptical or a spiral galaxy with lations are found to be rather independent _of the exact value
moderate star formation activity [W (O II) \ 16 A ]. The of q, except for very small values (q > 0.1 Gyr). We have
object 14.1103 (z \ 0.21) is detected 0at 15 km but not at calculated the total L (see Table 5) in the CFRS
2800on 143 galaxies that have been
radio wavelengths, and its SED is consistent with that of a 1415]52 Ðeld. It is based
local H II region. This is in agreement with the absence of an spectroscopically identiÐed and are representative of the
old stellar component as noted by Tresse et al. (1993) from 558 I \ 22.5 galaxies in the Ðeld (see Lilly et al. 1995b).
the very large Ha equivalent width (greater than 2500 A ). A Converted to a luminosity density, this value is in excellent
HST image (Fig. 6) shows an unresolved object, giving a agreement (to within 6%) with that found for the whole
diameter less than 250 pc. We are probably witnessing here CFRS Ðeld (Lilly et al. 1996). This is consistent with the
TABLE 4
PROPERTIES OF THE NONSTELLAR ISOCAM LW3 AND RADIO SOURCES WITH REDSHIFT
W (O II) Spectral Angular Morphological

0
CFRS z (A ) D3538 F2 F a F b L c L c SFR d Class SED Classe af Sizeg Final class ClassiÐcation
7 km 15 km 5 Ghz 2800 IR IR
Sources detected at 15 km and 5 GHz
14.0227 . . . 0.772 0.0 0.16 \150 274 \16 1.54 31.5 ^ 7.2 54.6 ^ 20.6 Quies SB]S ... ... SB]S Eh
14.0272 . . . 0.668 14.3 0.20 \150 415 \16 1.54 35.9 ^ 8.6 60.3 ^ 11.4 S]A SB]S ... ... SB]S Ch
14.0663 . . . 0.743 4.8 0.27 \150 275 20 0.99 127.5 ^ 4.8 213.1 ^ 17.8 S]A SBH 0.8 ^ 0.3 1.1 SBH Ch
14.0779 . . . 0.578 31.3 0.27 \150 357 \16 0.22 39.7 ^ 0.8 63.3 ^ 2.5 S]A Seyfert 2 /SBH ... ... SBH Ch
14.0818 . . . 0.899 22.1 0.20 \150 231 \38 2.13 50.2 ^ 5.9 86.6 ^ 10.1 S]A SB]S 0.7 ^ 0.2 ... SB]S Ch
14.0820 . . . 0.976 16.4 0.19 \150 284 33 0.03 12.7 ^ 1.0 ... S]A S 1.0 ^ 0.2 \1.3 S ...
14.0846 . . . 0.989 43.0 0.24 \150 350 \16 1.15 28.9 ^ 2.2 49.4 ^ 10.3 S]A SB]S/Seyfert 2 ... ... SB]S/Seyfert 2 Int
14.0909 . . . 0.978 18.7 0.34 \150 347 \16 0.39 46.0 ^ 40. ... S]A Seyfert 2 ... ... Seyfert 2 Ch
14.0998 . . . 0.430 11.8 0.43 \150 343 \26 0.20 5.7 ^ 0.4 10.4 ^ 5.2 S]A SB]S 0.7 ^ 0.3 ... SB]S Eh
14.1028 . . . 0.988 18.6 0.22 \150 413 31 0.61 205.0 ^ 25.7 ... S]A Seyfert 2/SB]S [0.2 ^ 0.3 20 Seyfert 2 SO/a
14.1042 . . . 0.875 17.0 0.23 176 231 \16 0.28 10.1 ^ 0.8 17.3 ^ 3.0 S]A SB]S/Seyfert 2 ... ... SB]S/Seyfert 2 SO/a
14.1103 . . . 0.209 0.0 ... \150 329 \16 0.03 1.3 ^ 0.3 H II H II ... ... H II C
14.1129 . . . 0.831 ... ... 207 292 \16 1.38 26.5 ^ 2.6 45.4 ^ 13.5 ... S]SB/Seyfert 2 ... ... S]SB Int
14.1139 . . . 0.660 19.3 0.23 161 787 79 1.38 157.5 ^ 0.5 280.6 ^ 1.8 S]A SBH 0.5 ^ 0.1 1.9 SBH Int
14.1190 . . . 0.754 16.0 0.30 \150 259 24 0.87 121.2 ^ 7.9 215.9 ^ 32.3 S]A SBH 0.8 ^ 0.2 1.6 SBH Sab
14.1302 . . . 0.548 19.7 0.14 \150 263 \16 0.86 6.5 ^ 2.5 È S Seyfert 1 ... ... Seyfert 1 Sh
14.1329 . . . 0.375 13.5 0.18 153 485 70 0.26 62.8 ^ 3.1 112.5 ^ 10.8 S]A SBH 0.8 ^ 0.1 0.9 SBH SO
14.1567 . . . 0.479 0.0 [0.02 \150 642 \16 4.19 6.5 ^ 2.5 È Seyfert 1 Seyfert 1 ... ... Seyfert 1 ...
14.9154 . . . 0.812 38.7 0.34 230 425 39 0.64 130.1 ^ 11.3 È S]A Seyfert 2 0.3^0.2 1.5 Seyfert 2 ...
Radio sources not detected at 15 km

14.0276 . . . 0.746 ... 0.32 \150 \250 37 0.57 16.2 ^ 1.6 È E SB]S \[0.4 10 Seyfert 2 ...
14.0573 . . . 0.010 ... ... 174 \250 37 0.57 11.1 ^ 0.5 ... Seyfert 1 Seyfert 1 [0.8 ^ 0.7 5]3 Seyfert 1 E
14.0727 . . . 0.463 32.2 0.28 \150 \250 14 0.54 11.3 ^ 2.8 19.4 ^ 3.3 S]A SB]S 1.0 ^ 0.3 1.3 SB]S ...
14.0854 . . . 0.992 0.0 0.13 \150 \250 35 0.08 26.3 ^ 3.9 44.9 ^ 7.1 Quies SB]S/Seyfert 2 0.0 ^ 0.3 \1.3 SB]S/Seyfert 2 SO/a
14.0937 . . . 0.838 7.6 0.04 \150 \250 1311 0.39 131.3 ^ 16.3 È S]A Seyfert 2 [0.1 ^ 0.1 \0.2 Seyfert 2 E
14.1041 . . . 0.372 56.0 0.30 \150 \250 47 0.28 8.0 ^ 3.2 È S]A Seyfert 2/SB]S [0.4 ^ 0.4 15]10 Seyfert 2 Irr
14.1303 . . . 0.985 ... ... \150 \250 45 22.20 115.7 ^ 34.7 QSO QSO 0.5 ^ 0.2 \1.6 QSO ...
14.9025 . . . 0.155 ... ... 243 \250 47 0.12 5.8 ^ 2.8 Liner Liner [0.4 ^ 0.4 15 ] 10 Liner Scd
a Fluxes corrected by aperture e†ects in kJy.

b Flux at 5 GHz in kJy from Fomalont et al. 1992.
c 2800 and IR (8È1000 km) luminosities in 1010 L .
_
d IR star formation rate in M yr~1.
_
e SED classiÐcation. See Table 3.
f Radio spectral index, between 1.5 and 5 GHz, from Fomalont et al. 1992.
g Angular size in arcseconds from Fomalont et al. 1992.
h From Schade et al. 1996.
0 These should be compared with their 2800 A luminosities

that range from 4 ] 108 to 2 ] 1010 L and are not very
_
di†erent from those of galaxies not detected at 15 km and
radio wavelengths. Thus, taking into account only the UV
luminosity of the 15 km and radio galaxies that are lumi-
nous, infrared sources would lead to severe underestimates
of their actual star formation rates. Nearly half of the global
-5
infrared luminosity of 15 km and radio galaxies is coming
from seven sources (four SBHs and three Seyfert 2). These
are detected at both wavelengths, except for the Seyfert 2
galaxy 14.0937. Estimates of infrared luminosities of these
bright galaxies should thus not be a†ected strongly by
source count biases or modeling uncertainties.
-10 Accounting for the global luminosity is slightly compli-
cated by the current observational status of the data.
Hammer et al. (1995) and recent follow-up work have nearly
completed the redshift identiÐcations of the radio kJy
sources. Among the 45 I ¹ 22.5 15 km sources that are
AB
not detected at radio wavelengths, only 16 possess a redshift
from the CFRS. The latter have not been selected a priori
-15
8 10 12 14 for their infrared properties and can be considered as repre-
sentative of the whole sample. In the calculation of the
global luminosity, we have assumed three di†erent scaling
FIG. 10.ÈComparison of the SEDs of Ðve SBH galaxies ( Ðlled circles,
factors : 1 for luminous galaxies detected in radio and at 15
separated by arbitrary vertical shifts) with a local averaged SBH SED km, 29/21 for galaxies detected only at radio wavelength
(dashed line ; from Schmitt et al. 1998). Fluxes at 60 km (open circles) are and 45/16, for the galaxies, with an additional statistical
derived from radio Ñuxes according to the radio-FIR correlation (S \ error of 25%. Table 5 presents the di†erent values of the
125S 60km
; Franceschini et al. 1994). Superposed to the Ðt are vertical bars
5Ghz global luminosities at 2800 A and in the infrared after
that display at each wavelength the standard deviation of the local tem-
plate. The bottom panel presents the standard deviation of the Ðve SBHs. excluding powerful AGNs (QSO and Seyfert 1). At IR wave-
lengths, Seyfert 2 contribute to more than a third of the
global luminosity.
Poisson error of 8% for the 143 representative galaxies.
5.2.3. Global IR L uminosity from Galaxies Undetected at 15 km
5.2. Global IR L uminosity and Radio
5.2.1. Interpolated IR L uminosities We consider here the subsample of the 49 galaxies that
Infrared (8È1000 km) luminosities are interpolated from possess a redshift, are not detected at 15 km or radio wave-
our MIR and radio Ñux measurements by using the tem- lengths, and have been observed at K. This subsample can
plate from Schmitt et al. (1997) with the smallest s2 for each be taken as representative of the 489 I ¹ 22.5 galaxies not
object. The template gives 5È7 points within the wavelength AB
detected at 15 km or radio wavelengths in the CFRS
range of interest, and errors can be estimated for each object 1515]52 Ðeld. For these galaxies, we use BV IK photo-
using our Monte Carlo simulations. Our quoted errors take metry and their 15 km and radio upper limits to investigate
into account Ñux measurements errors (13% on average) as their SEDs. We found that 19 of them have SEDs consistent
well as uncertainties in our classiÐcation scheme (5% on with those of local template SEDs for SBH galaxies and 16
average). The interpolated infrared luminosity for each have local template SEDs for SBL galaxies. The relatively
object, along with its associated error, is given in Table 4. large fraction of SBLs is compatible with the fact that they
are not detected in the MIR range. Other objects have their
5.2.2. Global IR L uminosity from Galaxies of the 15 km and SEDs consistent with those of local template SEDs for E/S
Radio Samples galaxies (nine objects) or AGN galaxies (Ðve objects). The
Galaxies detected at 15 km and/or radio wavelengths fraction of AGNs is in reasonable agreement with Hammer
have infrared luminosities in the range of (5 ] 1010)È et al. (1997), who Ðnd that D8% of all galaxies in the Ðeld at
(2 ] 1012) L (Table 4), comparable with those of typical z D 0.5 are Seyfert 2 galaxies, as indicated by their emission-
_ and Seyfert 2 galaxies (Genzel et al. 1998).
local starbursts line ratios. Figure 14 displays four examples of Ðts for these
TABLE 5
GLOBAL LUMINOSITIES [1010 \ L ] OF THE CFRS 1415]52 FIELD
_
15 km-RADIO NO DETECTED 15 km-RADIO CFRS 1415]52
L L L L L L
2800 IR 2800 IR 2800 IR
Upper limit . . . . . . . . . . . . . . . . . . . . 29 ^ 6 1778 ^ 607 217 ^ 29 2161 ^ 214 246 ^ 35 3939 ^ 821
AGN contribution . . . . . . . . . . . . 4^1 680 ^ 291 17 ^ 2 84 ^ 27 21 ^ 3 724 ^ 318
““ Old star ÏÏ contribution . . . . . . 1^0 115 ^ 28 25 ^ 3 44 ^ 16 26 ^ 3 159 ^ 44
Lower limit . . . . . . . . . . . . . . . . . . . . 24 ^ 5 984 ^ 288 175 ^ 24 511 ^ 71 199 ^ 29 1495 ^ 359
-5
-10
-15
8 10 12 14 8 10 12 14
FIG. 11.ÈSED of galaxies classiÐed as S]SB, separated by arbitrary vertical shifts, compared with SEDs of local spiral galaxies with a starburst
component (S]SB ; dashed line). The arrow indicates the radio Ñux limit (16 kJy) for some galaxies. Galaxies detected only at radio wavelengths are marked
with an asterisk.
objects. All objects but four have been reasonably classiÐed

by this method. The four remaining objects have Ðnally
been classiÐed as starbursts.
5.2.4. IR L uminosity Density Related to Star Formation and

Uncertainties
The luminosity density of star-forming galaxies may be
simply estimated by summing up the contribution of all
starburst galaxies (SBH, SBL, and S]SBH) identiÐed in the
Ðeld. Uncertainties, which include errors in source counts,
Ñux errors, and uncertainties in the identiÐcation, would
come from our Monte Carlo simulations.
However, several complications could a†ect the validity
of such a calculation. First, the origin of the UV and infra-
red light from Seyfert 2 galaxies is unclear and can be attrib-
uted either to the AGN, star formation, or both. Indeed,
Seyfert 2 galaxies show a wide range in their infrared
properties, and few color diagrams are able to distinguish
them from starbursts. Contamination of the luminosity
density by AGNs is potentially the most serious problem
3000 3500 4000 4500 5000 for any estimate of the star formation density. Second,
cirrus can contribute to the infrared luminosity, and this is
likely the case for the objects classiÐed as S]SBH. It is
FIG. 12.ÈAverage spectrum of the Ðve SBH and S]SBH galaxies with
z \ 0.7. Spectral features are indicated. The average spectrum has been
often assumed that the 60 km luminosity of apparently
computed by using median values to account for luminosity variation from quiescent early-type galaxies is related to cirrus excited by
one object to another. the emission of the underlying stellar population (Sauvage
-5
-10
-15
-20
8 10 12 14 8 10 12 14
FIG. 13.ÈSED of AGNs detected with ISO at 15 km compared with local SEDs. (left) All objects are compared with a Seyfert 2 SED template, except for
14.9025 (Liner SED template). (right) Powerful AGN (QSO and Seyfert 1) SEDs compared with power-law or radio-quiet QSO (Sanders et al. 1989). Galaxies
detected only at radio wavelengths are marked with an asterisk and their Ñuxes at 15 km by an arrow, assuming a upper limiting Ñux of 250 kJy.
& Thuan 1992). This is however disputed by Devereux & 3. The infrared luminosities of S]SBH galaxies are
Hameed (1997), who present some counterexamples and assumed to be only coming from star formation.
argue that young stars can be an important source of the
FIR light of early-type galaxies. One may wonder, however, The lower limit is obtained with the following assump-
about the exact nature of early-type galaxies with a signiÐ- tions :
cant population of young stars embedded in dust. Might 1. All objects not detected at 15 km and radio wave-
they not be equivalent to our S]SBH galaxies ? In that lengths have zero extinction, and hence the star formation
case, the cirrus contribution to the infrared luminosity of rate estimated at infrared wavelengths equals that estimated
these galaxies cannot be signiÐcant, as the starburst com- at UV wavelengths.
ponent contributes to most of their bolometric luminosities. 2. The 2800 A and 60 km luminosities of AGNs and
Recall that former estimates of the SFR co-moving quiescent (E) galaxies are assumed to be related, respec-
density from the UV light are uncertain because : tively, to the active nucleus and the old stellar component.
1. They include contributions at 2800 A from AGNs and 3. The infrared luminosities of S]SBH galaxies are
old stellar populations. assumed to be related to both cirrus and star formation, in
2. They do not account for the UV light reprocessed by the proportion of the ratio of the bolometric luminosities of
dust into FIR radiation. the two components.
We wish here to take into account the deÐciencies men- Table 5 displays the global luminosities in the CFRS
tioned above and calculate upper and lower limits to the 1415]52 Ðeld as derived from UV (2800 A ) and infrared
luminosity density related to star formation. The upper limit (8È1000 km) Ñuxes for the sample of objects detected either
is obtained with the following assumptions : at 15 km or radio wavelengths and that containing objects
1. The 15 km and radio Ñuxes of all objects not detected not detected at both wavelengths. The AGN contributions
at those wavelengths are set equal to their detection limit. are given in the upper limit case. They can attain non-
This provides an upper limit to their infrared Ñux. negligible values (D20%), including at UV wavelengths.
2. The 2800 A and 60 km luminosities of AGN-classiÐed This represents a major uncertainty for the SF densities
objects are assumed to be powered by star formation. derived from UV Ñuxes. The errors take into account uncer-
FIG. 14.ÈFour examples of the SED of galaxies that are not detected at 15 km and radio wavelengths but possess B[K observations.
tainties in source counts, Ñux measurements errors, and 6.2. Star Formation Density Derived from the IR L ight
uncertainties in the classiÐcation scheme. Density
The SFR can also be inferred from infrared luminosities,
assuming that they result mostly from dust heating by
6. ESTIMATION OF THE COSMIC STAR FORMATION RATE
young stars in the optically thick limit. The SFR calibration
6.1. Star Formation Density Derived from the UV L ight depends mainly on the burst duration and IMF slope at
Density both high- and intermediate-mass ranges. Following Ken-
The star formation history has been estimated by various nicutt (1998), we adopt the same IMF as that assumed to
authors (see Madau et al. 1996 ; Lilly et al. 1996 ; Hammer et calculate the SFR derived from UV light, and
al. 1997) on the basis of the CFRS and/or HDF surveys SFR \ 1.71 ] 10~10(L /L ) . (4)
IR IR _
using star formation rates derived from 2800 A or [O II] The above equation is based on the models of Lehnert &
3727
Ñuxes. Since the UV light is dominated by emission from Heckman (1996) for a continuous burst with age 10È100
more or less massive stars, the total SFR as derived only Myr. Condon (1992) calibrated the SFR from the non-
from the Ñux at 2800 A is somewhat uncertain, as the thermal radio luminosity (L ), assuming the Galactic rela-
extrapolation to the low-mass end (M ¹ 5 M ) of the IMF NT
_ (1997) have tion between L and the radio supernova rate. He then
is not constrained. Moreover, Hammer et al. NT
derived the SFR calibration from the FIR-radio corre-
suggested that dust and metallicity can severely a†ect these IR D2 of the UV-to-optical emission is
lation, showing that
estimates and mask the true evolution of the cosmic star re-emitted between 40 and 3 120 km (see also Helou et al.
formation rate. We adopt in the calculation of the star for- 1988). After rescaling his value by assuming a Salpeter IMF
mation rate (SFR) from the UV light the calibration of (0.1È100 M ), Condon (1992) obtains
Madau et al. (1998) and Kennicutt (1998), assuming a Salpe- _
SFR \ 2.22 ] 10~10(L /L ) . (5)
ter (1955) IMF with mass limits 0.1 and 100 M : IR IR _
_ The di†erence between the two authors is at the 28%
SFR \ 5.045 ] 10~10(L /L ) . (3) level, in agreement with the Kennicutt uncertainty estimate
2800 2800 _
of ^30% on his calibration, after comparison with other

models (Hunter et al. 1986 ; Lehnert & Heckman 1996 ;
Meurer et al. 1997).
The key parameter in these estimations is the star forma-
tion rate of massive stars (typically º5 M ), which are
_
mainly responsible for the UV continuum, the nonthermal
radio continuum, and FIR luminosities reradiated by dust.
The SFR /SFR ratio calculated in the following is essen-
IR UV of the IMF slope and in particular of any
tially independent
extrapolation toward low stellar masses.
6.3. Star Formation Density Missed by UV Observations
We aim to calculate here the cosmic SFR from the infra-
red emission and compare it with that derived from the UV
emission. We Ðrst consider the ““ directly observed ÏÏ SFR,
based on observations of the I ¹ 22.5 galaxies, as present-
AB
ed by Lilly et al. (1996). In ° 5.1 we have calculated the total
L contributed by all galaxies with I ¹ 22.5 and z ¹ 1.
2800 AB
In ° 5.2 we have computed the total L from those galaxies
IR
with either S º 250 or S º 16 kJy and from CFRS
15km 5GHz
galaxies not detected at 15 km and radio wavelengths.
We have calculated the SFR /SFR ratio, accounting
for all galaxies not classiÐed as IR powerful
2800 AGNs. It ranges
from 2.5 ^ 0.95 in the lower limit case to 5.4 ^ 1.9, when FIG. 15.ÈMetal production and star formation history for z ¹ 1 (see
galaxies undetected at 15 km and in radio have their Ñuxes text). Our points ( Ðlled circles, labeled ISO-VLA-CFRS) are 2.9 times
higher in SFR density or in metal production than those (open circles)
at these wavelengths equal to the detection limit. These previously derived from the UV Ñux density at 2800 A . Other points are
values can, however, be a†ected by uncertainties on the from Gallego et al. (1995 ; open triangle), Connolly et al. (1997 ; open stars),
calibration of the SFR in the UV and FIR wavelength Madau et al. (1998 ; HDF ; open squares) and have not been corrected for
ranges. Nevertheless, the values we Ðnd for SFR and extinction.
SFR show some consistency, since for all galaxies2800 SFR is
IR than SFR IR
larger . These ratios give extinctions ranging
2800 L \ 196 ] 1010 L in the Ðeld. According to equation
from A \ 0.49 to 0.87 if we assume a standard galactic 2800
extinction
V law. If the true SFR density is given by the FIR (3), this corresponds to_ a global SFR of 980 M yr~1.
Adopting the Tresse & Maddox extinction 2800 _ the
estimate,
estimation, then from 35% to 85% of the global SFR co-
moving density for z ¹ 1 is not taken into account when global infrared luminosity would be equal to 1445 ] 1010
only the UV Ñux density is considered. L (equation 4). This is in very good agreement with the
_
average value in the fourth column of Table 5 and adds
Figure 15 shows the co-moving SFR density evolution
with look-back time (for H \ 50 km s~1 Mpc~3 and q \ more conÐdence in the VLA-ISO-CFRS average values
0.5). The VLA-ISO-CFRS 0points ( Ðlled dots) represent0the shown in Figure 15.
average values of the upper and lower limits deÐned above. These average values imply that the stellar mass formed,
Since there is no evidence for a change in the extinction for since z \ 1 would be 3.3 ] 108 M Mpc~3, slightly larger
_
z ¹ 0.5 as compared with 0.5 \ z ¹ 1, we have adopted the than the estimation of the present-day stellar mass of
same redshift bins as those used in previously deriving the 3 ] 108 M Mpc~3 (Cowie et al. 1996 ; Glazebrook et al.
_
SFR from the UV Ñuxes. VLA-ISO-CFRS data are 1995). Of course, the uncertainties on the SFR estimations
2.9 ^ 1.3 larger than former UV estimated values, even are large, and our lower estimate is still consistent with the
though the latter accounted for all sources, including strong present day stellar mass. However, we believe that this will
AGNs. The error bars account for incompleteness errors constitute a serious problem, when uncertainties will be
(Lilly et al. 1996) as well as uncertainties in the determi- lowered by larger and deeper surveys at infrared and radio
nation of the SFR density at IR wavelengths. wavelengths. The problem gets worse if one accounts for the
Error bars in Figure 15 are very large mostly because we additional star formation at z [ 1. There may be two ways
have assumed that the blue galaxies undetected at 15 km out. First, half of the SFR density comes from D10% of the
and radio wavelengths have no extinction (lower limit case) Ðeld galaxies, those detected at 15 km and radio wave-
or have Ñux densities equal to the detection limits (upper lengths. These objects with large SFR (especially those with
limit case), both of which are clearly unrealistic assump- greater than 100 M yr~1) may have IMFs that signiÐ-
_ Salpeter law. Alternatively, the local
tions. It is interesting to check whether our average value cantly deviate from the
makes sense. Starbursts detected by both ISOCAM and the stellar mass density may have been systematically under-
VLA are rare. They have large extinctions ranging from estimated.
A \ 1.5 to 2 and show red continua (Fig. 12). Tresse &
V
Maddox (1998) predict an extinction of 1 mag at 2800 A (or
7. COMPARISON WITH THE HDF/ISOCAM SURVEYS
A \ 0.48) for the blue Ðeld galaxy population at z ¹ 0.3. If The number densities of faint MIR and radio sources
V same value holds on average for blue galaxies up to found in the CFRS 1415]52 Ðeld and the HDF appear to
the
z \ 1, we can predict in principle the infrared luminosity of be a reasonable agreement. However, our results are not
the Ðeld galaxy population not detected at 15 km and radio consistent with those derived for the HDF by Rowan-
wavelengths. For them, Table 5 gives an average value Robinson et al. (1997). The HDF is 18.5 smaller in area than
the CFRS 1415]52 Ðeld, and from our results we would To derive FIR luminosities (and hence SFRs) requires
expect 0.3 starbursts with SFR larger than 100 M yr~1, as complete information from UV-to-radio wavelengths. It
compared with the four starbursts with SFRs from _ 500 to
seems reasonable to use templates based on well-known
1010 M yr~1 claimed by Rowan-Robinson et al. (1997). local objects, especially for the most powerful star-forming
_
No simple explanation can account for this discrepancy of galaxies.
more than 1 order of magnitude. The calibration of the
SFR by Rowan-Robinson et al. (1997) di†ers slightly 8. CONCLUSIONS
FIR
from ours (their proportionality factor in eq. [4] is Observations of distant Ðeld galaxies have been obtained
2.6 ] 10~10 instead of our 1.7 ] 10~10). We simply stress with ISOCAM down to unprecedented Ñux levels at 6.75
that : (S º 150 kJy) and 15 km (S º 250 kJy). We have attempted
1. Our source detection and identiÐcation procedures are to make our samples at both those wavelengths as complete
robust (see also Flores et al. 1998), while source identiÐca- as possible by a careful reduction of the data. That the
tions can be difficult in the HDF, since several galaxies in samples are complete is suggested by the good positional
the HDF can lie within a single ISOCAM pixel. correlation between source identiÐcations at 6.75 and 15 km
2. Rowan-Robinson et al. (1997) have used a three- and radio wavelengths.
component model (a stellar component giving rise to the Source densities are comparable at 6.75 km (1944
UV-to-MIR emission plus cirrus and starburst components, S [ 150 kJy sources deg~2 ; Flores et al. 1998), 15 km (2808
giving rise to the MIR-to-radio emission) to Ðt galaxy S [ 250 kJy sources deg~2 ; this paper), and 5 GHz (1440
SEDs. This type of model has several degrees of freedom, S [ 16 kJy sources deg~2 ; Fomalont et al. 1991). Star-
including the relative energy ratio between the di†erent forming objects contribute, respectively, 50%, 73%, and
components. The number of free parameters is larger than 26% of the extragalactic counts at 6.75 and 15 km and 5
the number of available data points (1È3) in the MIR-to- GHz. This suggests that the 60 km luminosity density is
radio wavelength range. strongly dominated by star-forming galaxies. The fraction
3. Among the twelve HDF galaxies whose SEDs were of z [ 1 objects is found to be less than 32%, 43%, and 40%
Ðtted by Rowan-Robinson et al. (1997), Ðve have limits on of the extragalactic counts at 6.75 and 15 km and 5 GHz,
their radio luminosities that are signiÐcantly lower than the respectively. The 15 km survey is found to be rather efficient
starburst model predictions (see their Fig. 1). in selecting high-redshift objects, since sources with I ¹
AB as
Instead of force-Ðtting the data points by multiple- 22.5 and S º 250 kJy have a median redshift of 0.76,
15km
parameter starburst models, we propose here a more con- compared with 0.59 for the whole CFRS.
servative empirical approach of Ðtting SEDs of galaxies The 15 km survey, combined with radio and optical data,
with z ¹ 1 with SED templates of well-studied local gal- allows us to identify the most powerful star-forming objects
axies (see Schmitt et al. 1997). The Ðts are generally quite (with SFRs larger than 100 M yr~1) in the Ðeld, at least up
_ are found among the 578
good, especially for starburst (SB) SEDs that are consistent to z \ 1. Four such objects (0.7%)
with the standard radio-FIR correlation. A pure starburst I ¹ 22.5 galaxies, and they contribute to 18% of the SFR
has its energy distribution dominated mostly by star forma- density. From their UV or [O II] emission-line proper-
3727
tion. Its SED includes massive star light with or without ties, these objects cannot be distinguished from galaxies
extinction at UV wavelengths, re-emission of the absorbed with more modest rates of star formation. If we correct for
UV light at MIR and FIR wavelengths, and starburst extinction, assuming that their optical spectra from 2500È
thermal emission and synchrotron radiation from super- 4000 A is reddened by the standard galactic extinction
nova remnants at radio wavelengths. Locally, Schmitt et al. curve (with A derived from the SFR /SFR ratio), they
would appearVas young starbursts with
IR a moderate
UV
(1997) show that starbursts, when compared with normal popu-
spirals or ellipticals, have a smaller spread in their SEDs lation of A stars [W (Hd) \ 3.5 A in the combined
from FIR to radio wavelengths. There is no evidence that spectrum]. The 15 km 0sample contains highly reddened
these properties change signiÐcantly in the redshift range young starbursts as well as a larger number of galaxies
0 ¹ z ¹ 1. (S]SB) with lower SFRs, which all contain a signiÐcant
Strong starbursts (with SFRs larger than 70 M yr~1) population of A stars (S]A galaxies). This is consistent
_ such
should be detected by deep radio surveys to kJy levels, with the scenario of strong starburst episodes followed by
as those by Fomalont et al. (1991) in the CFRS 1415]52 the last phases of the burst, where the IR emission is still
Ðeld or Fomalont et al. (1997) in the HDF. The fact that high because of dust heating by intermediate-mass stars
these are not detected in the HDF at radio wavelengths (M \ 1È3 M ; Lisenfeld, Volk, & Xu 1997).
_
casts some doubt on the large SFRs derived by Rowan- The combination of 15 km and radio samples probably
Robinson et al. (1997) for objects without radio counter- gives a good representation of the galaxy population in a
parts. As for the three radio sources with S º 16 kJy very deep 60 km survey. Interpolated infrared (8È1000 km)
and z ¹ 1 that are found in the HDF (Fomalont 5GHzet al. 1997), luminosities based on Ðts of galaxy SEDs by local templates
one (ISOHDF 12 36 46 ] 62 14 06, z \ 0.96) is found by from radio-to-UV wavelengths imply that 75% ([40%,
Rowan-Robinson et al. (1997) to have the highest SFR ]10%) of the star formation rate density for z ¹ 1 is hidden
(1010 M yr~1) in their sample. It is at z \ 0.96, has an by dust. No evidence has been found for an evolution of
elliptical _
morphology, is undetected at 15 km (as most of the that fraction in the above redshift range. The global opacity
elliptical galaxies of the Hammer et al. 1995 radio sample) of the universe up to z \ 1 ranges from A \ 0.5 to 0.85. A
and shows highly ionized emission typical of a Seyfert subsample of 16 15 km galaxies observed Vby the HST indi-
galaxy (Mg II j2799, [Ne V] j3426, and [Ne III] j3868). It is cates that more than a third of the star formation hidden by
likely a Seyfert 1 galaxy, since the Mg II line is broad and its dust is associated with interacting galaxies or mergers.
radio emission shows some variability (Richards et al. 1998). Estimates of the SFR from UV Ñuxes carry some uncer-
It is doubtful that its emission is powered by star formation. tainties because of two main reasons : one related to the UV
light reprocessed by dust into FIR radiation and the other when the corresponding stellar mass formed, since z \ 1 is
to the probably important contribution of AGN light at compared with the present-day stellar mass density. This
UV wavelengths. In spite of the small statistics of the could raise an important question about the universality of
sample considered here, our work allows a Ðrst glimpse of the IMF, especially in the high-SFR galaxies detected by
the true SFR density for z ¹ 1. From our careful data ISOCAM and the VLA and described in this paper. Studies
analysis and the use of multiwavelength data from radio-to- of other CFRS Ðelds with the same multiwavelength tech-
UV, we believe that the true SFR density lies within the nique are needed to improve the source statistics, determine
region delimited by the still large error bars in Figure 15. more accurately the SFR density, and study in more detail
Our average value, 70% of the star formation rate density its redshift evolution.
hidden by dust, is consistent with a 1 mag absorption at
2800 A for the blue galaxy population (see Tresse &
Maddox 1998) and less than 4% of the galaxies being highly We thank Marc Sauvage for useful discussions. We are
reddened, detected at 15 km and radio wavelengths, and also grateful to David Schade, who made available the HST
contributing to 50% of the global IR luminosity density. images of the CFRS 1415]52 Ðeld in stamp format. Com-
Although our values are lower by a factor 2.8 than those ments and criticism from an anonymous referee have led us
of Rowan-Robinson et al. (1997), they might be too high to greatly improve the manuscript.
REFERENCES
Brinchmann, J., et al. 1997, ApJ, 499, 112 Kennicutt, R. 1992, ApJ, 388, 310
Bruzual, G., & Charlot, S. 1995 Kennicutt, R., Jr. 1998, ApJ, 498, 541
Cesarsky, C. J., et al. 1996, A&A, 315, L32 Kessler, M. F., et al. 1996, A&A, 315, L27
ÈÈÈ. 1998, in preparation Laurant, O., & Mirabel, F. 1998, in preparation
Clements, D., et al. 1998, preprint (astro-ph/9809054) Lehnert, M., & Heckman, T. 1996, ApJ, 472, 546
Condon, J. 1992, ARA&A, 30, 575 Leitherer, M., Fergunson, H., & Heckman, T. 1995, A&AS, 186, 3004
Condon, J., & Roderick, J. 1988, AJ, 96, 30 Lilly, S., Le Fèvre, O., Crampton, D., Hammer, F., & Tresse, L. 1995a, ApJ,
Connolly, A., Szalay, A., Dickinson, M., SubbaRao, M., & Brunne, R. 1997, 455, 50
ApJ, 486, L11 Lilly, S., Le Fèvre, O., Hammer, F., & Crampton, D. 1996, ApJ, 460, L1
Cowie, L., Hu, E., Songaila, A., & Cohen, J. 1996, ApJ, 112, 839 Lilly, S. J., Hammer, F., Le Fèvre, O., & Crampton, D. 1995b, ApJ, 455, 75
Desert, F. X., et al. 1998, A&A, in press Lisenfeld, U., Volk, H., & Xu, C. 1996, preprint (astro-ph/9605118)
Devereux, N., & Hameed, S. 1997, AJ, 113, 599 Lutz, D., et al. 1998, ApJ, 505, 103
Elbaz, D., et al. 1998, preprint (astro-ph/9807209) Madau, P., et al. 1996, MNRAS, 283, 1388
Flores, H., et al. 1998, A&A, in press Madau, P., Pozzetti, L., & Dickinson, M. 1998, ApJ, 498, 106
Fomalont, E. B., et al. 1997, ApJ, 475, L5 Meurer, G., et al. 1997, AJ, 114, 54
Fomalont, E. B., Windhorst, R. A., Kristian, J. A., & Kellerman, K. I. 1991, Moshir, M. 1989, IRAS Faint Source Survey, Explanatory Supp. v. 1
AJ, 102, 1258 (Pasadena : CalTech)
Franceschini, A., et al. 1991, A&AS, 89, 285 Puget, J., et al. 1998, in preparation
Franceschini, A., Mazzei, P., Zotti, G., & Danesse, L. 1994, ApJ, 427, 140 Richards, E., et al. 1998, AJ, 116, 1039
Gallego, J., Zamorano, J., Aragon-Salamanca, A., & Rego, M. 1995, ApJ, Rowan-Robinson, M., et al. 1997, MNRAS, 289, 490
455, 1 Salpeter, E. 1955, ApJ, 121, 161
Genzel, R., et al. 1997, preprint (astro-ph/9711255) Sanders, D., et al. 1988, ApJ, 328, 35
Glazebrook, K., et al. 1998, preprint (astro-ph/9808276) ÈÈÈ. 1989, ApJ, 347, 29
Glazebrook, K., Peacock, J., Miller, L., & Collins, C. 1995, MNRAS, 275, Sauvage, M., & Thuan, T. X. 1992, ApJ, 396, L69
169 Schade, D., Crampton, D., Hammer, F., Le Fèvre, O., & Lilly, S. J. 1996,
Groth, E., et al. 1994, BAAS, 185, 5309 MNRAS, 278, 95
Hammer, F., Crampton, D., Lilly, S., Le Fèvre, O., & Kenet, T. 1995, Schmitt, H., Kinney, A., Calzetti, D., & Storchi-Bergmann, T. 1997, AJ,
MNRAS, 276, 1085 114, 592
Hammer, F., et al. 1997, ApJ, 480, 59 Spinoglio, L., et al. 1995, ApJ, 453, 616
Hammer, F., Schade, D., Crampton, D., Lilly, S., & Le Fèvre, O. 1996, in Spinoglio, L., & Malkan, M. 1989, ESA/IRSA 343
Science with the Hubble Space T elescope II, ed. P. Benvenuti et al. Tresse, L., Hammer, F., Le Fevre, O., & Proust, D. 1993, A&A, 277, 53
(Baltimore : STScI), 101 Tresse, L., & Maddox, S. J. 1998, ApJ, 495, 691
Helou, G., Khan, I., Malek, L., & Boehmer, L. 1988, ApJS, 68, 151 Wrobel, J., & Heeschen, D. 1988, ApJ, 335, 677
Hunter, J., Sandford, M., Whitaker, R., & Klein, R. 1986, ApJ, 305, 309

Flores 1999 ApJ 517 148

Uploaded by

Copyright:

Available Formats

Flores 1999 ApJ 517 148

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Flores 1999 ApJ 517 148

Uploaded by

Copyright:

Available Formats

THE ASTROPHYSICAL JOURNAL, 517 : 148È167, 1999 May 20

15 MICRON INFRARED SPACE OBSERV AT ORY 1 OBSERVATIONS OF THE 1415]52 CANADA-FRANCE

ISO a d CFRS za I b V b K b dc Pd Fluxe Error

Catalog 1 : Objects with S/N [ 4, P \ 0.02

Catalog 2 : Objects with S/N [ 4, P(d, I) [ 0.02

Catalog 3 : Objects with S/N [ 4, without optical counterpart in I

Catalog 4 : Objects with 4 [ S/N [ 3, P \ 0.02

ISO a d CFRS za I b V b K b dc Pd Fluxe Error

Catalog 5 : Objects with 4 [ S/N [ 3, P [ 0.02

250 Most (71%) of the 15 km optical counterparts are classi-

10 3.3. Morphologies from HST and CFHT

4. SPECTRAL ENERGY DISTRIBUTIONS AND OBJECT

E Sp LINER Seyfert 2 SBL SBH Hybrid

14.1103 14.1157 14.1329 14.0998

14.1139 14.1190 14.1129 14.1042

14.1028 14.0846 14.1166 14.0198

14.0302 14.0711 14.1554 14.1232

available for most objects, except those detected only at

describes the current mix of old and young stellar popu-

W (O II) Spectral Angular Morphological

Radio sources not detected at 15 km

a Fluxes corrected by aperture e†ects in kJy.

0 These should be compared with their 2800 A luminosities

objects. All objects but four have been reasonably classiÐed

5.2.4. IR L uminosity Density Related to Star Formation and

of ^30% on his calibration, after comparison with other

You might also like

0 These should be compared with their 2800 A luminosities