Near-infrared faint galaxies in the Subaru Deep Field: Comparing the theory with observations for galaxy counts, colors, and size distributions to K similar to 24.5

Galaxy counts in the K band, (J-K) colors, and apparent size distributions of faint galaxies in the Subaru Deep Field (SDF) down to K similar to 24.5 were studied in detail. Special attention has been paid to take into accoun t various selection effects, including the cosmological dimming of surface brightness, to avoid any systematic bias that may be the origin of controve rsy in previously published results. We also tried to be very careful about systematic model uncertainties; we present a comprehensive survey of these systematic uncertainties and dependence on various parameters, and we have shown that the dominant factors to determine galaxy counts in this band ar e cosmology and number evolution. We found that the pure luminosity evoluti on (PLE) model is very consistent with all the SDF data down to K similar t o 22.5, without any evidence for number or size evolution in a low-density, Lambda -dominated flat universe, which is now favored by various cosmologi cal observations. On the other hand, a number evolution of galaxies with et a similar to 2, when invoked as the luminosity conserving mergers as phi* p roportional to (1 + z)(eta) and L* proportional to (1 + z)(-eta) for all ty pes of galaxies, is necessary to explain the data in the Einstein-de Sitter universe. If the popular Lambda -dominated universe is taken for granted, our result then gives a strong constraint on the number evolution of giant elliptical or early-type galaxies to z similar to 1-2 that must be met by a ny models in the hierarchically clustering universe, since such galaxies ar e the dominant population in this magnitude range (K less than or similar t o 22.5). A number evolution with eta similar to 1 is already difficult to r econcile with the data in this universe. On the other hand, number evolutio n of late-type galaxies and/or dwarf galaxies, which has been suggested by previous studies of optical galaxies, is allowed from the data. In the fain ter magnitude range of K greater than or similar to 22.5, we found a slight excess of observed counts over the prediction of the PLE model when ellipt ical galaxies are treated as a single population. We suggest that this disc repancy reflects some number evolution of dwarf galaxies and/or the distinc t populations of giant and dwarf elliptical galaxies which have been known for local elliptical galaxies.