The Lyman break photometric ``drop-out'' technique was first utilized for the detection of high redshift quasars, but was then much more finely tuned as a detection technique for high redshift galaxies by Steidel and his collaborators in the mid-1990s. The technique relies on the large break in the continuum flux from an object that occurs at the 912 Å Lyman limit from neutral hydrogen absorption in the line-of-sight. Multi-band images of a field containing high redshift galaxies can be used to identify those objects that have very red colors as a result of the redshifted Lyman limit falling between any two filters. The technique is refined by having multiple filters that also can detect the smaller break at L (1216 Å ), and the rather blue continuum longwards of L.
As first used extensively by Steidel et al. (1996) the technique was used to detect z ~ 3 objects by their lack of flux in the U-band (hence the descriptor ``U-band drop-outs''). The technique was often applied by using three or four filters and defining a region in the two-color plane in which such ``drop-outs'' were most likely to occur. Steidel and his collaborators used UGR ground-based images, while, for the HDF, the four band images allowed the use of a plane that was essentially (U - B) vs (V - I). Examples of the two-color plane and the selection function are given also in Dickinson (1998), as is a very instructive visual representation of the ``drop-out'' technique for a galaxy at z = 3.
This technique has proved to be remarkably useful for detecting z > 2 galaxies with ground (and space) imaging in the ``optical'' (~ 0.3-1 µmm). A key advantage of this technique is that it is essentially free of selection effects, with little contamination from low redshift objects. All high redshift objects above a given magnitude limit will be detected, provided the S/N is high enough in all the bands, particularly the bluest band where an upper limit must be established. Occasionally red stellar objects and dusty galaxies contaminate the sample, but the fraction is small with high S/N images.
The only objects which might still be detectable spectroscopically, but would be missed by this technique are those with a very strong emission line and very weak continuum fluxes (typically the strong line would be Ly). Stockton (this volume) and Hu et al. (1999) discuss examples of such objects, and also discuss narrow-band imaging searches for detecting such objects (see also Section 8 below).