We describe the design, manufacture, and performance of bare-fiber integral field units (IFUs) for the SDSS-IV survey MaNGA (Mapping Nearby Galaxies at APO) on the the Sloan 2.5 m telescope at Apache Point Observatory (APO). MaNGA is a luminosity-sel
ected integral-field spectroscopic survey of 10,000 local galaxies covering 360-1030 nm at R ~ 2200. The IFUs have hexagonal dense packing of fibers with packing regularity of 3 um (RMS), and throughput of 96+/-0.5% from 350 nm to 1 um in the lab. Their sizes range from 19 to 127 fibers (3-7 hexagonal layers) using Polymicro FBP 120:132:150 um core:clad:buffer fibers to reach a fill fraction of 56%. High throughput (and low focal-ratio degradation) is achieved by maintaining the fiber cladding and buffer intact, ensuring excellent surface polish, and applying a multi-layer AR coating of the input and output surfaces. In operations on-sky, the IFUs show only an additional 2.3% FRD-related variability in throughput despite repeated mechanical stressing during plate plugging (however other losses are present). The IFUs achieve on-sky throughput 5% above the single-fiber feeds used in SDSS-III/BOSS, attributable to equivalent performance compared to single fibers and additional gains from the AR coating. The manufacturing process is geared toward mass-production of high-multiplex systems. The low-stress process involves a precision ferrule with hexagonal inner shape designed to lead inserted fibers to settle in a dense hexagonal pattern. The ferrule inner diameter is tapered at progressively shallower angles toward its tip and the final 2 mm are straight and only a few um larger than necessary to hold the desired number of fibers. This process scales to accommodate other fiber sizes and to IFUs with substantially larger fiber count. (Abridged)
We present a new analysis of stellar mass functions (MF) in the COSMOS field to fainter limits than has been previously probed to z~1. Neither the total nor the passive or star-forming MF can be well fit with a single Schechter function once one prob
es below 3e9 Msun. We observe a dip or plateau at masses ~1e10 Msun, and an upturn towards a steep faint-end slope of -1.7 at lower mass at any z<1. This bimodal nature of the MF is not solely a result of the blue/red dichotomy. The blue MF is by itself bimodal at z~1. This suggests a new dichotomy in galaxy formation that predates the appearance of the red sequence. We propose two interpretations for this bimodality. If the gas fraction increases towards lower mass, galaxies with M_baryon~1e10 Msun would shift to lower stellar masses, creating the observed dip. This would indicate a change in star formation efficiency, perhaps linked to supernovae feedback becoming much more efficient. Therefore, we investigate whether the dip is present in the baryonic (stars+gas) MF. Alternatively, the dip could be created by an enhancement of the galaxy assembly rate at ~1e11 Msun, a phenomenon that naturally arises if the baryon fraction peaks at M_halo ~1e12 Msun. In this scenario, galaxies occupying the bump around M* would be identified with central galaxies and the second fainter component having a steep faint-end slope with satellites. While the dip is apparent in the total MF at any z, it appears to shift from the blue to red population, likely as a result of transforming high-mass blue galaxies into red ones. At the same time, we detect a drastic upturn in the number of low-mass red galaxies. Their increase with time reflects a decrease in the number of blue systems and so we tentatively associate them with satellite dwarf galaxies that have undergone quenching.
