Keck spectroscopy of faint radio sources has identified a population of galaxies which are expected to produce a large fraction of the elusive submillimeter background. The sources have a median flux at 850 microns of 1.3 mJy, and thus are too faint to be detected with submm instruments like SCUBA on the JCMT. However these galaxies are still extremely luminous, and are very difficult to identify from optical surveys for star-forming galaxies.

LEFT PANEL: a sample Keck-LRIS spectrum of a redshift 2.2 radio galaxy, showing a star-forming spectral type.
RIGHT PANEL: the redshift distributions as a function of 850micron flux (bottom), and the integral contributions to the submm background (top)
Red circles are submm galaxies -- SMGs (the 65% fraction of the blank field sources which are detected in the radio). Green triangles are radio galaxies -- RGs, undetected in the submm, and having non-AGN spectra. Submm flux has been inferred by taking the weighted submm sum (1.3mJy+-0.4mJy) and picking the single dust temperature that makes the radio luminosities produce 1.3mJy
Upper panel: shows the cumulative contribution to the submm background (with "1" being the total Fixsen et al. (1996) background).
RED and GREEN curves: SMGs and RGs corrected for spectroscopic incompleteness only (assuming that the complete radio source population down to 20muJy has properties as probed by these ~30 non-SMM RG sources) MAGENTA curve: SMGs+RGs
DARK BLUE curve: LBGs from Adelberger & Steidel (2000), LBG-inferred 850micron count
CYAN curve: LBGs - RGs + SMGs (assuming that these RGs would typically be in LBG samples, with VLA-radio being an efficient way to find `1mJy' sources).

  The submm background probes the total unobscured star formation and AGN history of the Universe, emphasizing the formative period in the first half of the Universe (z=10 to 1). Contrast the far-IR (200 micron) background which also probes the unobscured energetic history of the Universe, but is dominated by the latter half of the history from z=1 to the present (z=0). The galaxies which dominate the submm background are likely to be star-forming galaxies with typical star formation rates of ~50 Msun/yr, at redshifts greater than one. Despite their importance to galaxy evolution, virtually nothing is known about these galaxies.

  The redshifts, temperatures, and clustering of submm galaxies (SMGs) are now well understood (respectively, Chapman et al. 2003,2004; Kovacs et al. 2004; Blain et al. 2004), realizing a complete characterization of the population. The submm is also the natural place to study these ~1 mJy galaxies which dominate the background, however current and upcoming instruments (SCUBA, MAMBO, BOLOCAM, LABOCA) are confusion limited before reaching these SFRs - they only detect the extreme and rare tail of the luminosity function (and it's hard to pinpoint them in optical images).

  These `1mJy' galaxies (~50 Msun/yr) are the brightest of the z=1-3 "Lyman-break galaxy" luminosity function, but ironically are very difficult to identify in optical surveys for star-forming galaxies. The reason for this is a strong correlation between bolometric luminosity and degree of obscuration by dust (Adelberger & Steidel 2000); the most intrinsically luminous galaxies are not the galaxies with the largest apparent brightness.

  Using Keck-LRIS, we have measured spectroscopic redshifts for thirty objects likely to be representative of this class of galaxy. These galaxies are microJansky radio sources with faint optical counterparts lying at redshifts around two, with UV spectra clearly showing star-forming signatures. These galaxies are undetected in the submm, placing a strong upper limit on their luminosities. The key to the result is that these galaxies do not show AGN characteristics, thus the radio is very likely to correlate directly with the submm emission. We are able to demonstrate that the space density of these objects represents 40% of the 1mJy submm galaxies, and therefore ~20% of the total submm background.