THE LAST OF FIRST: THE FINAL CATALOG AND SOURCE IDENTIFICATIONS

The FIRST survey, begun over 20 years ago, provides the definitive high-resolution map of the radio sky. This Very Large Telescope (VLA) survey reaches a detection sensitivity of 1 mJy at 20 cm over a final footprint of 10,575 deg(2) that is largely coincident with the Sloan Digital Sky Survey (SDSS) area. Both the images and a catalog containing 946,432 sources are available through the FIRST Web site (http://sundog.stsci.edu). We record here the authoritative survey history, including hardware and software changes that affect the catalog's reliability and completeness. In particular, we use recent observations taken with the JVLA to test various aspects of the survey data (astrometry, CLEAN bias, and the flux density scale). We describe a new, sophisticated algorithm for flagging potential sidelobes in this snapshot survey, and show that fewer than 10% of the cataloged objects are likely sidelobes, and that these are heavily concentrated at low flux densities and in the vicinity of bright sources, as expected. We also report a comparison of the survey with the NRAO VLA Sky Survey (NVSS), as well as a match of the FIRST catalog to the SDSS and Two Micron Sky Survey (2MASS) sky surveys. The NVSS match shows very good consistency in flux density scale and astrometry between the two surveys. The matches with 2MASS and SDSS indicate a systematic similar to 10-20 mas astrometric error with respect to the optical reference frame in all VLA data that has disappeared with the advent of the JVLA. We demonstrate strikingly different behavior between the radiomatches to stellar objects and to galaxies in the optical and IR surveys reflecting the different radio populations present over the flux density range 1-1000 mJy. As the radio flux density declines, stellar counterparts (quasars) get redder and fainter, while galaxies get brighter and have colors that initially redden but then turn bluer near the FIRST detection limit. Implications for future radio sky surveys are also briefly discussed. In particular, we show that for radio source identification at faint optical magnitudes, high angular resolution observations are essential, and cannot be sacrificed in exchange for high signal-to-noise data. The value of a JVLA survey as a complement to Square Kilometer Array precursor surveys is briefly discussed.