1. INTRODUCTION

In September 1985 the faint multiaperture imaging spectrograph PUMA was operated for the first time at the Cassegrain focus of the Canada-France-Hawaii-Telescope (CFHT) to study the dynamical state of distant clusters of galaxies and to detect QSOs which are weakly lensed and magnified by rich clusters (Fort et al. 1986). A strange ring-like structure appeared on all the CCD images of the distant cluster of galaxies A370 (z = 0.374) obtained in the R and V bands. The arc was definitely not an artifact and a paper was submitted to A&A early in 1986 to draw attention to this exotic structure (Soucail et al. 1987). Independently, Lynds & Petrosian (1986) reported in the Bulletin of the American Astronomical Society the discovery of arcs in A370, A2218 and Cl2244-02. In fact, astronomical archives contained several photographs, vidicon images and CCD images where arcs in A370 and A963 (Butcher, Oemler & Wells 1983) could be seen, and Hoag (1981) already reported seeing elongated structures in A370. Lynds & Petrosian (1989) reviewed some previous historical observations of arc-like structures. Nobody seems to have started systematic surveys to search for such arcs, probably because it was commonly thought that the cores of the richest clusters could not reach a projected surface matter density large enough to split images of distant sources (see however Nottale & Hammer 1984; Turner, Ostriker & Gott 1984, Webster 1985). The attention of observers was instead focussed on the problem of the dynamical and photochemical evolution of clusters of galaxies. Incidentally before the advent of CCDs, the very faint surface brightness of these arcs made them almost unobservable. At the time of the re-discovery of arcs, few astronomers were ready to accept the gravitational lens hypothesis. The Toulouse group itself had almost discarded the hypothesis because of the absence of a counter image as would be expected for a simple lens with circular symmetry. The ideas were rather to find an explanation with events occurring at the cluster redshift such as: cooling flows (Soucail et al. 1987a, b), bow shocks (Begelman & Blandford 1987), or light echo models (Katz 1987, Milgrom 1987, Katz & Jackson 1988). It was actually Paczynski (1987) who first realized that the location of arcs in the cores of dense clusters, their circular geometry, their blue colors and faint surface brightnesses were all tremendous pieces of evidence that we were observing the first gravitationally distorted images of distant galaxies located behind the clusters. Almost immediately Kovner (1987a) argued that cluster cores can act as lenses and showed that it was possible to reproduce the single highly magnified arcs in A370 and Cl2244 with an elliptical potential (Kovner 1988).

Figure 1. CCD image of the giant arc discovered in A370. The arc extends over 25 arcsec, and is actually a merging of three images of a background source at z = 0.725. The other galaxies are cluster members at redshift z = 0.375. A370 appears as a bimodal cluster dominated by two very luminous galaxies. One is visible at the top of the image, while the other one lies close to the arc. Note also the faint pairs (B2-B3), (C1-C2) and (D1-D2), which are multiple images of other background sources (Kneib et al. 1993; see Sect. 2). The notation (B2-B3) come from Fort et al (1988). Kneib et al. predicted that B4 must be the third image associated with (B2-B3).

Independently, Hammer (1987) presented a good modeling of the arc in A370 with a spherical halo plus a distribution of point masses associated with individual galaxies. This fascinating event initiated an intense observational effort to obtain spectra of the arcs. Soucail et al. (1988) succeeded in getting good spectra with curved slits of several pieces of the A370 arc with the faint object spectrograph EFOSC/PUMA2 on the 3.6 meter telescope of the European Southern Observatory. This measurement was not easy because the arc has a mean surface brightness B = 24.6 per square arcsecond, and it is also contaminated by several cluster members (Fig. 1). A strong emission line ([OII] 3727) and several other spectral features were detected along the arc. The spectrum corresponds to an Sa/Sb galaxy at a redshift z = 0.724, nearly twice the redshift of the lensing cluster (Fig. 2). Fortunately, it was possible to confirm this redshift with spectra obtained by Miller & Goodrich (1988), and Lynds & Petrosian (1988). The gravitational lens hypothesis was at least verified without any possible doubt for A370.

Figure 2. Spectrum of the giant arc discovered in A370, obtained at the 3.60m ESO telescope with the EFOSC/PUMA2 faint object spectrograph. The [OII] 3727 Å line is redshifted by a factor 1.725 and is detected all along the giant arc, which demonstrates that we are observing a strongly distorted background galaxy at about twice the distance of the cluster.

At the same time an important observational event occurred in the arcs story. Very deep CCD photometry of 12 selected blank fields at high galactic latitude revealed a large population of faint extended objects uniformly distributed on the sky (Tyson 1988). Their blue colors, magnitudes and surface brightnesses supported the idea that they were distant galaxies at a redshift larger than 1 rather than a nearby population of faint dwarf galaxies. Their number density up to a limiting magnitude B_j of 27 was about 100 per square arc minute. If located at large distance, these galaxies constituted a dense grid of randomly distributed sources that should be distorted by cluster-lenses (see the simulations by Grossman & Narayan 1988). This effect was suspected in A2218 (Pelló et al. 1988) and immediately verified in A370 (Fort et al. 1988, Fig. 1 and 2). The cluster was indeed revealing a few faint distorted objects with an orthoradial orientation with respect to the cluster center: the so-called arclets. They had a blue color and a surface brightness similar to Tyson's population. This was the last additional evidence in favor of the gravitational hypothesis for the luminous arc in A370.

Following the earlier prediction of Webster (1985), it was understood at this moment that there may exist a new and abundant class of gravitational lenses which would be a new tool for cosmological studies (Grossman & Narayan 1988, Turner 1988, Nottale 1988, Fort 1989, Fort et al. 1993): the cluster cross section is much larger than for galaxies acting as lenses, and the density of background sources is much higher than quasars and therefore can map properly the typical angular scale of distant clusters. Contrary to point-like sources, arclets are magnified galaxies which show an observable distortion from which we can infer the shear. Furthermore, a large number of very rich and distant clusters of galaxies were known which should produce many gravitational images of distant galaxies! Finally, the observation of a large excess of tangentially aligned objects in a very deep image of the cluster A1689 by Tyson, Valdes & Wenk (1990) convinced the astronomical community to support systematic observations of arcs. The observation of very faint arc(let)s was a challenge strongly motivated by the idea of probing the distribution of dark matter in galaxy clusters via their weak lensing effects on the distant background galaxies (Tyson et al. 1990). After several years of observations and the first successful modeling of cluster-lenses having multiple arc systems (Kassiola, Kovner & Fort 1992; Mellier, Fort & Kneib 1993; Kneib et al. 1993), most of the initial hopes are becoming reality. The potential of some lenses is now so well determined at the cluster center that the location of secondary images of arcs can be predicted in advance and found by observers afterward. With good modeling of the lens there is a possibility of determining the probable redshift of many arclets from an analysis of their distortion and without any spectroscopic measurements! Clusters of galaxies as very large natural gravitational telescopes allow us to study the distribution and evolution of distant field galaxies which would have hardly been observable without the gravitational magnification effect. There is also hope of finding multiple image systems which can give constraints on the geometry of the universe and the value of the cosmological constant. It is remarkable to think that almost all of the topics of this review were in the fantastic vision of F. Zwicky (1937), who foresaw the possible cosmological interest of gravitational images formed by massive nebulae.

In this paper we present the status of this new observational area. The review is organized as follows: section 2 gives the basic concepts of gravitational lensing in order to understand the formation of arcs in rich galaxy clusters. For this part the reader who is not familiar with the field of gravitational lensing can also read introductions to the basic concepts of gravitational lensing by Blandford & Narayan (1992). More comprehensive recent reviews on the theoretical aspects of gravitational optics can be found in papers by Blandford & Kochanek (1988), Blandford & Narayan (1986), and particularly the large monograph by Schneider, Ehlers & Falco (1992) with complete references therein. Some other theoretical aspects related to individual observational issues for the modeling of arc(let)s will be quoted in the following chapters. Section 3 will present a rapid overview of the observations, which were already reviewed in Fort (1989), Soucail (1992), and Fort (1992). The technical difficulties of observing arcs will be also briefly discussed. Section 4 is devoted to the distribution of dark matter in clusters of galaxies. The discussion concerns both the modeling of the potential in the region of multiple images (strong lensing) and the use of arclets for a direct measurement of the gravitational shear of the background population of galaxies at large radius from the cluster center. Several theoretical approaches have been proposed since the pioneering observational work of Tyson et al. (1990), but only a few observational attempts have been made due to small sizes of current CCDs. This section is long compared with the others because most of the scientific return from cluster lenses obtained during the last 5 years concerns these points. Section 5 is related to the study of the distribution and evolution of very distant field galaxies using galaxy clusters as gravitational telescopes, and the possibility of observing a new class of highly magnified distant objects. Section 6 reviews what could be done in the near future for cosmography. This part is more tentative since the observations are not yet completed. Discussion of the possible evolution of the observations with new generation telescopes like HST, ROSAT, and VLTs will be given in the conclusion.