Target Selection and Validation of DESI Quasars

TL;DR

This work presents the DESI quasar target selection strategy, combining optical g,r,z photometry with infrared W1,W2 data to robustly separate QSOs from stars. A Random Forest classifier, trained on spectroscopically confirmed QSOs and unlabeled stars, yields a main selection of 16.5<r<23.0 with PSF morphology and region-specific probability thresholds, delivering over 200 QSOs deg^-2 (including ~60 at z>2.1) at a target density of 310 deg^-2. Automated quasar cataloging using Redrock, Mg II, and QuasarNet achieves purity >99% and efficiency ~93–99% across conditions, validated against visually inspected data and SDSS catalogs. The results show stable QSO yields with exposure time and across the focal plane, minimal systematics after masking, and redshift distributions in good agreement with QLF predictions, indicating strong readiness for DESI clustering and Ly-α forest science.

Abstract

The Dark Energy Spectroscopic Instrument (DESI) survey will measure large-scale structures using quasars as direct tracers of dark matter in the redshift range 0.9<z<2.1 and using Ly-alpha forests in quasar spectra at z>2.1. We present several methods to select candidate quasars for DESI, using input photometric imaging in three optical bands (g, r, z) from the DESI Legacy Imaging Surveys and two infrared bands (W1, W2) from the Wide-field Infrared Explorer (WISE). These methods were extensively tested during the Survey Validation of DESI. In this paper, we report on the results obtained with the different methods and present the selection we optimized for the DESI main survey. The final quasar target selection is based on a Random Forest algorithm and selects quasars in the magnitude range 16.5<r<23. Visual selection of ultra-deep observations indicates that the main selection consists of 71% quasars, 16% galaxies, 6% stars and 7% inconclusive spectra. Using the spectra based on this selection, we build an automated quasar catalog that achieves a >99% purity for a nominal effective exposure time of ~1000s. With a 310 per sq. deg. target density, the main selection allows DESI to select more than 200 QSOs per sq. deg. (including 60 quasars with z>2.1), exceeding the project requirements by 20%. The redshift distribution of the selected quasars is in excellent agreement with quasar luminosity function predictions.

Figures (21)

• Figure 1: Distribution of the PSF Depth $r$ in the DR9 Legacy Imaging surveys footprint. The $r$-band is used to define the magnitude limit for DESI QSO target selection. The solid black line shows the Galactic plane. Three different imaging footprints are highlighted. The blue region is the combination of BASS and MzLS (designated as North region in the paper). The red region is the DES part of DECaLS. The green region, which excludes the red and the blue regions, is the non-DES part of DECaLS. The union of red and the green regions is named as South region in the paper.
• Figure 2: Colors in the optical or near-infrared of objects photometrically classified as stars (red) or spectroscopically classified as QSOs (from blue to yellow dots, depending on their redshift). The color $grz - W$ allows us to reject stars based on the "infrared excess" of QSOs.
• Figure 3: Density map of the DR9 QSO target selection. The solid black and dashed blue lines show respectively the Galactic plane and the plane of the Sagittarius Stream.
• Figure 4: Relative QSO target density in the North, South (Non-DES) and South (DES) regions as a function of each observational parameter (see Sec. \ref{['sec:parameters']} for the definition of the parameters). The relative QSO target density is a mean value after rejecting outliers. The histograms represent the distributions of each observational parameter in the three regions. The color code is blue, green and red, respectively, for the North, South (Non-DES) and South (DES) regions.
• Figure 5: Relative $\chi^2$ difference between extended and 'PSF' models as a function of the $\chi^2$ difference, for COSMOS/HST objects. The violet dots correspond to objects confirmed as point-like sources in HST imaging. The green dots correspond to objects identified as extended galaxies in the HST imaging. The blue dots are HST point-like sources that are classified as extended objects in the DECaLS DR9 photometric catalogs.