1 Introduction

Gravitational lensing by clusters of galaxies is recognized as a powerful cosmological tool. By measuring the various properties of the gravitationally magnified and distorted mirages of background galaxies (the so-called arcs and arclets), we can learn important information about both the lensing cluster and the faint, distant, lensed background galaxies. In principle, cluster lenses may provide a means to study normal galaxies at high redshift (); galaxies that would be out of reach of even our largest telescopes if not for the magnification of the lensing clusters. These galaxies are likely to be free from selection effects since they lie serendipitously behind the foreground clusters and are not selected because they are intrinsically bright, or are luminous radio or X-ray emitters. The majority of the arcs and arclets discovered so far are very blue, and are believed to come from the ubiquitous background population of faint blue galaxies (FBGs) seen in deep field galaxy surveys (Tyson 1988; Colless et al.1990 and 1993; Lilly, Cowie & Gardner 1991; Koo & Kron 1992; Lilly et al.1995 among others). It may be possible to use gravitational lensing to constrain the redshift range of the FBGs by examining the frequency of lensing as a function of cluster redshift (i.e., Tyson et al. 1990; Miralda-Escudè 1993b; Smail et al. 1994). But for all its potential as a probe of distant galaxies, cluster gravitational lensing is presently finding its most useful application as a tool for studying the clusters themselves. Cluster lens systems are relatively common compared to lensed QSOs because the high space density of the faint blue galaxies provides a convenient background grid whose distortion we can observe around any sufficiently massive and concentrated foreground cluster. Lensing can be used to measure the cluster mass, independent of assumptions about virialization of the galaxies or hydrostatic equilibrium of the hot X-ray-emitting intracluster gas. Moreover, gravitational lensing provides a means to explore the actual shape of a cluster's total mass distribution, both visible and dark matter, at small radii (kpc) using strong gravitational lensing (giant arcs), and at large radii using weak gravitational lensing (Miralda-Escudè 1991 & 1992; Kaiser 1992; Kaiser & Squires 1993; Mellier et al.1994; Broadhurst et al. 1994; Smail et al. 1994; Kaiser et al. 1996; Schneider 1996; Squires et al.1996a and 1996b; Squires et al.1997; Geiger & Schneider 1998; Clowe et al.1998).

Clearly, a large, homogeneous sample of clusters of galaxies spanning a broad range of redshift would be extremely valuable. Previous attempts to investigate the statistics or frequency of lensing have relied on optically-selected clusters with all of the problems inherent in the optical selection (see Lynds & Petrosian 1989; Smail et al.1991). It is well known that projection and contamination problems can be avoided if clusters are selected based on their primary baryonic constituent--the hot, intracluster gas that is a copius emitter of X-rays--rather than on the optical galaxies which are merely trace components. We entered this investigation with two primary goals: first, to discover new cluster lens systems, and second, to systematically study a complete sample of distant clusters whose selection criteria were well defined and designed to avoid the contamination and superposition problems. The sample was selected from the Einstein Observatory Extended Medium Sensitivity Survey (EMSS) list of clusters given by Gioia et al.(1990a).

LeFèvre et al.(1994) presented preliminary results from an analysis of lensing in a subsample of the 16 most X-ray luminous and distant EMSS clusters (z$\,\gt\,$0.2 and $L_{\rm x}$$\,\gt\,$$4 \ 10^{44}$ ergs^-1). Note, however, that this subsample was extracted from the list of Henry et al.(1992) and did not include the 3 added clusters with z$\,\gt\,$0.5 (MS1137+66, MS0451-03, MS1054-03; Gioia & Luppino 1994) since their redshifts were not known at the time. For the most part, the LeFèvre et al. data consisted of V and I "snapshots'' taken with the CFHT that allowed identification of the brightest arcs that might be present in these clusters. From this preliminary subsample, they concluded that bright arcs with large axis ratios (l/w >>1) were fairly common in X-ray selected clusters, and that this high frequency of lensing implied that the mass density profiles of the lensing clusters must be compact ($r_{\rm c}$ < 100 kpc; see also Hammer et al.1993; Wu & Hammer 1993; Hammer et al.1997).

In this paper, we present the results from the 38-cluster sample. In Sect. 2, we discuss the properties of the EMSS galaxy cluster sample and our selection criteria for this survey. We describe the optical observations and data reduction and calibration in Sect. 3. In Sect. 4, Sect. 5 and Sect. 6 we present our data and describe in detail those clusters that contain giant arcs. In addition to the obvious lensing cases, we also describe those systems with candidate arcs whose lensing origin needs follow-up confirmation, either with deeper imaging or spectroscopy. In Sect. 7, we investigate the lensing fraction as a function of both redshift and X-ray luminosity. We also discuss the constraints the EMSS cluster lenses place on the cluster mass density profiles. Finally, in Sect. 8, we summarize our results and discuss the prospects for future work.

Throughout this paper, we assume a cosmology where H₀=50km s^-1 Mpc^-1, $q_0\,=0.5$, but we occasionally use both conventional parameterizations of the Hubble constant, $h_{50} \equiv H_0/50$ and $h \equiv H_0/100$,depending on the context of the discussion.