Big BOSS Spectral Simulations Nick Mostek UCB DOE

Big. BOSS Spectral Simulations Nick Mostek (UCB) DOE Review of Big. BOSS, Dec. 6 -7, 2011

Outline • Overview of spectral simulator purpose • Description of simulated components — Input source and sky spectra — Imaging profiles at focal plane and detector surface — Instrument throughput — Sky subtraction • Results from a few case studies • R&D trade studies • Other ongoing simulation efforts • Summary Nick Mostek Talk B 4. 5 Simulation 2

Big. BOSS Spectral Simulations BBspecsim is an end-to-end simulation of the Big. BOSS instrument • Includes astrophysical sources, atmospheric effects, instrument throughputs, PSFs, and detector characteristics. • Used to evaluate system performance, including exposure time estimates and redshift success rates • Supported ETC projections at community workshop • Software will be essential to R&D trade studies in the upcoming year • Led by Nick Mostek (UCB) with contributions from David Schlegel, Alex Kim (LBNL), and Arjun Dey (NOAO). New developments to come from Stephen Bailey (LBNL), Kyle Dawson, and Adam Bolton (Utah). Nick Mostek Talk B 4. 5 Simulation 3

Astrophysical Sources • Generates ELG templates [OII] — [OII] line flux and line velocity — Constant line ratios for other lines (Hb, [OIII], Ha, Lya) Hb [OIII] Ha • Generates flat mag. AB spectrum • Flexible flux spectrum input z=1. 4 Nick Mostek Talk B 4. 5 Simulation — Simple text input format — Includes example LRG and QSOs — Options to scale spectrum to an alternative SDSS magnitude or redshift 4

Atmospheric Effects Lunar Phase • • • Atmospheric transmission from KPNO extinction curve + high resolution measurements from the Mc. Math Solar FTS in the absorption bands Sky emission taken from UVES catalog (R=45, 000) and a lunar phase model (Krisciunas and Schaefer, 1991) reflecting the solar spectrum UVES sky brightness matches BOSS sky in the continuum Nick Mostek Talk B 4. 5 Simulation 5

Mayall Delivered Image Quality Median ‘Best’ DIQ Median DIQ Data measured from 17, 268 (KPNO) and 47, 266 (CTIO) images from MOSAIC cameras in NOAO Science Archive (from Dey & Valdes 2012) Wavelength (nm) Similar seeing achieved at Prime Focus for both Blanco and Mayall Nick Mostek Talk B 4. 5 Simulation 6

Focal Plane PSF + • + = Convolve imaging profile on focal plane assuming: — Seeing (Moffat profile with Beta factor) — Field-weighted, wavelength-dependent optical blur — Exponential Sersic source profile • Calculate fraction of light injected into fiber of an input size (1. 5” default) • Also include losses due to pointing centroid offsets / astrometric error Nick Mostek Talk B 4. 5 Simulation 7

Spectrum generator • Additional throughputs and efficiencies for system components — Includes mirrors, fibers, VPH gratings, optics, and detectors — Use test data when available, otherwise simulated from standardized software • Each spectrograph arm has a changing optical spot profile in both the dispersion and spatial direction • We perform a 2 D interpolation between ZEMAX-generated spot profiles — Currently only implemented in the dispersion direction (single spectrum) Red arm – 7800 A Red arm – 10300 A Nick Mostek Talk B 4. 5 Simulation 8

Sky Subtraction Sky Lines Observed Spectrum • Sky subtracted assuming a sky-only observation with the RMS error on the mean reduced by sqrt(nsky)=sqrt(25)=5 • Perform an extraction of spectrum assuming collapsed 1 D spot and fitting F=A·X for every spatial column (spaxel) of the 2 D spectrum • Signal and error computed for each 1 D pixel from fit A scalar value and 1 fit error • Future simulation will generate 2 D PSF and use full 2 D extraction algorithm (c. f. Bolton Talk) Mean 25 sky fibers Sky-Subtracted Spectrum l Nick Mostek Talk B 4. 5 Simulation [OII] 9

Redshift Detection Studies Credit: Alex Kim • • Simulated 10 k spectra at random redshift and [OII] flux in the red spectrograph arm Fit data with emission line templates with zfind (SDSS legacy code) Regions of decreased redshift efficiency are from bright sky lines and the z>1. 6 detector cutoff Can translate results to selected ELG n(z) distribution and project the detected sample Nick Mostek Talk B 4. 5 Simulation 10

Redshift Detection Studies Detected redshifts from 0. 5<z<1. 6 All ELG targets Successful ELG redshifts • ELG detected n(z) after folding in [OII] detection efficiency (blue histograms) • Detected [OII] flux distribution shows most failures occur at low line flux (S/N limited) • Can also project margin on spectrograph efficiency or [OII] luminosity function - 20% reduction in [OII] = 14% total n(z) loss - 50% reduction in [OII] = 40% total n(z) loss - Largest n(z) loss is at z~1 Nick Mostek Talk B 4. 5 Simulation 11

Signal to Noise Studies • Instrument continuum sensitivity can be calculated for a specific AB magnitude • Sensitivity calculations based on mean values of measured distributions (source size, seeing, airmass, etc. ) in dark sky conditions • Similar to BOSS sensitivity after rescaling for mirror size, but needs further cross-checks during R&D Nick Mostek Talk B 4. 5 Simulation 12

Resolution Studies R=2000 Single line fit to Dc 2=1 Compared to correct fit 7 s Det. R=4000 Single line fit to Dc 2=9 Compared to correct fit • • Trade studies are underway to set resolution requirements that minimize catastrophic redshift errors Focus on splitting the [OII] doublet at R>4000, reduced c 2 is separable from a broad single line Nick Mostek Talk B 4. 5 Simulation 7 s Det. 13

BBspecsim R&D Projects Future Improvements: • • Full CCD images 2 D spectral extraction algorithm Simulation cross-checks Support future workshops (targeting, instrumentation, and community) Trade studies: • • Resolution requirements vs catastrophic redshift error rates Redshift detection efficiencies for LRGs and QSOs Optimized exposure times vs survey speed 2 D PSF algorithm and calibration requirements Instrumentation optimizations: • • • Fiber size Variable losses due to positioner / astrometric error Spectrograph and grating efficiencies Nick Mostek Talk B 4. 5 Simulation 14

R&D Survey Strategy • • Simulates the entire 5 year survey using a set sky tiling and distributions of seeing, lunar brightness, and weather Tuned to acquire lowest airmass observations Developed by Knut Olsen and Arjun Dey (NOAO) Further survey planning R&D studies to be performed by Kyle Dawson (Utah) Nick Mostek Talk B 4. 5 Simulation 15

R&D Survey Strategy • Trade studies of tile density, target completeness, and fiber allocation efficiency • For 5000 fiber in 3 deg Fo. V, we expect ~80% target completeness and ~80% fiber efficiency for average target densities of ~3000 targets deg-2 • Developed by Mike Blanton and Pat Jelinsky. More sophisticated allocation schemes are being studied by Marco Azzaro (Spain) Target galaxy Fiber patrol area 1. 0 Baseline Fiber Efficiency 0. 8 4 deg 2 Fo. V 0. 6 3 deg 2 Fo. V 0. 4 0. 2 2 deg 2 Fo. V 0. 0 1 3 5 7 Tile Visits Nick Mostek Talk B 4. 5 Simulation 9 11 13 16

Summary BBspecsim is a software tool developed to study Big. BOSS spectra • • Flexible input for S/N calculations on various sources Extensive efforts to either measure or model end-to-end throughput Telescope and spectrograph imaging PSFs are incorporated into the sim Capable of Monte Carlo simulations for instrument trade studies — Testing redshift detection rates — Resolution Requirements — Exposure time calculations Upcoming R&D and improvements • • • Update to new spectrograph design and optical prescription Full up simulation of spectral image along spatial (slit) direction Incorporate full 2 D spectral extraction algorithm Nick Mostek Talk B 4. 5 Simulation 17