COLORS AND K-BAND COUNTS OF EXTREMELY FAINT FIELD GALAXIES

We combine deep K-band (W. M. Keck Telescope) with V- and I-band (New Technology Telescope) observations of two ''blank'' high Galactic lati tude fields, surveying a total of similar to 2 arcmin(2). The K-band n umber-magnitude counts continue to rise above K approximate to 22 mag, reaching surface densities of few x 10(5) deg(-2). The slope for the galaxy counts is approximately d log (N)/d mag deg(-2) = 0.23 +/- 0.02 over the range 18-23 mag. While this slope is consistent with other r ecent deep K-band surveys, there is a definite scatter in the normaliz ations by about a factor of 2. In particular, our normalization is sim ilar to 2x greater than the galaxy counts reported by Djorgovski et al . in 1995. Optical-near-infrared color-magnitude and color-color diagr ams for all objects detected in the V + I + K image are plotted and di scussed in the context of grids of Bruzual-Charlot isochrone synthesis galaxy evolutionary models. The colors of most of the observed galaxi es are consistent with a population drawn from a broad redshift distri bution. A few galaxies at K approximate to 19-20 are red in both color s (V-I greater than or similar to 3; I-K greater than or similar to 2, consistent with being early-type galaxies having undergone a burst of star formation at z greater than or similar to 5 and viewed at z simi lar to 1. At K greater than or similar to 20, we find several (approxi mately eight) ''red outlier'' galaxies with I-K greater than or simila r to 4 and V-I less than or similar to 2.5, whose colors are difficult to mimic by a single evolving or nonevolving stellar population at an y redshift unless they either have quite low metallicity or are highly reddened. We compare the data against the evolutionary tracks of seco nd-burst ellipticals and against a grid of models that does not constr ain galaxy ages to a particular formation redshift. The red outliers' surface density is several per square arcminute, which is so high that they are probably common objects of low luminosity L < L(). Whether these are low-metallicity, dusty dwarf galaxies, or old galaxies at hi gh redshift, they are curious and merit spectroscopic follow-up.