1 1 0 Se p 20 01 Spectral Reduction Software for the DEEP 2 Redshift Survey

The DEEP2/DEIMOS redshift survey, which will begin observing in the Spring of 2002, will gather high quality spectra on ∼ 60000 galaxies in order to study the evolution of the properties and large scale clustering of galaxies at z ∼ 1. The data rate from DEIMOS will be in excess of 1 Gbyte/hour, and it is therefore imperative to employ completely automated data reduction techniques to manage the analysis. We here describe aspects of our data pipeline, which will make extensive use of B-splines for the skysubtraction stage and for the combination of multiple frames. 1 Spectral Reduction Strategy The Keck II Deep Imaging Multi-Object Spectrograph (DEIMOS) is intended for imaging and multislit spectroscopy over a field of view that is approximately a rectangle of size 16’ by 5’ and over the wavelength range 0.42-1 μm. With its focal plane mask allowing slitlets spectroscopy for ∼ 100-150 objects, DEIMOS is optimally designed for large surveys of faint objects. The initial use of DEIMOS will be largely to undertake a massive redshift survey of ∼ 60,000 z ≈ 1 galaxies, the DEEP2 survey [1]. The technical complexity of the reductions and the shear volume of data force us to a rather new approach to spectral data reduction. Working closely with the SDSS team, we are developing a dedicated, fully automatic data reduction package based on the IDL codes of Schlegel, Burles, and Finkbeiner. 1.1 Spectral Tracing The DEIMOS mosaic imager is an 8K by 8K pixel camera composed of 8 individually mounted 2K by 4K CCDs manufactured by MIT Lincoln Laboratory. The pixel size is 15 μm. The DEIMOS CCDs are thick, high resistivity devices, with enhanced QE in the near-IR and much reduced fringing compared to thinner chips. In contrast to the VLT/VIMOS design, DEIMOS will have smaller multiplexing but many more spectral pixels per target, and there will not be multiple objects within a given row of data. The large number of pixels in the dispersion direction (8K) allows high resolution with substantial spectral range, i.e. a wide redshift interval coverage, while the spatial extension allow us to cover 16’ of sky and a large multiplexing capability. The spectrum of each object will be dispersed across 2 CCDs separated by a gap whose size is a polynomial function of the position along the spatial direction. We have developed a series of fast algorithms to trace the spectra across the 2 CCDs, align and rectify them and fit the gap size with sub pixel precision at each trace position. Based on the traces of the edges of the curved slitlet spectra over the CCD frame, the data for individual slitlets is shifted by whole pixels in the spatial direction to produce rectangular spectra. We next perform a non linear wavelength solution fit to calibrating arc lamps using the formula λ(x) = aj · L [x+G(y)θ( x 4096 +G(y) )] where L(u) are Legendre polynomials of order j, x is the pixel position along the dispersion direction, G(y) is the unknown gap at position y along the spatial direction and θ is the Heaviside theta function.