DESI 2024 II: Sample Definitions, Characteristics, and Two-point Clustering Statistics

TL;DR

DESI DR1 provides four large-scale tracers (QSO, LRG, ELG, BGS) with carefully constructed LSS catalogs and extensive randoms, masking, and weighting schemes to control survey systematics. The paper details the data processing pipeline, redshift selection, geometry veto masks, fiber-assignment completeness, imaging and spectroscopic systematics treatment, and the construction of 2-point statistics (both configuration- and Fourier-space) with window-function corrections. It validates the DR1 clustering measurements against realistic mocks (altmtl and FFA EZmocks), discusses covariance modeling, and presents the results and limitations of raw DR1 measurements, BAO analyses, and full-shape constraints. The work establishes a robust, publicly released dataset and methodology for large-scale structure studies, with insights into systematic uncertainties and guidance for future DESI data releases.

Abstract

We present the samples of galaxies and quasars used for DESI 2024 cosmological analyses, drawn from the DESI Data Release 1 (DR1). We describe the construction of large-scale structure (LSS) catalogs from these samples, which include matched sets of synthetic reference `randoms' and weights that account for variations in the observed density of the samples due to experimental design and varying instrument performance. We detail how we correct for variations in observational completeness, the input `target' densities due to imaging systematics, and the ability to confidently measure redshifts from DESI spectra. We then summarize how remaining uncertainties in the corrections can be translated to systematic uncertainties for particular analyses. We describe the weights added to maximize the signal-to-noise of DESI DR1 2-point clustering measurements. We detail measurement pipelines applied to the LSS catalogs that obtain 2-point clustering measurements in configuration and Fourier space. The resulting 2-point measurements depend on window functions and normalization constraints particular to each sample, and we present the corrections required to match models to the data. We compare the configuration- and Fourier-space 2-point clustering of the data samples to that recovered from simulations of DESI DR1 and find they are, generally, in statistical agreement to within 2\% in the inferred real-space over-density field. The LSS catalogs, 2-point measurements, and their covariance matrices will be released publicly with DESI DR1.

Figures (21)

• Figure 1: The left-hand panel displays the comoving number density for the four discrete galaxy and quasar tracers in the DESI DR1, estimated for a sample with no targeting incompleteness. Solid curves show the results for data targeted with DECam photometry ('S') and dashed curves for the target data that used BASS/MzLS photometry ('N'). The right-hand panel displays the same information, but without completeness corrections and with data limited to the redshift ranges adopted in DR1 cosmological analysis. In this panel the solid and dashed curves show results for the North and South galactic cap regions (NGC and SGC) respectively: the differences are driven primarily by the relative assignment completeness in each region. Note that the left and right panels use logarithmic and linear axis scaling, respectively, which allow different distinctive features to be seen. In both panels, verticle grid lines appear at the limits of the redshift bins used to define the final DESI DR1 samples; on the right, they appear at greater thickness than those otherwise at every 0.1 in redshift.
• Figure 2: The top two panels show the number of overlapping tiles in DR1 bright and dark time. The lower four panels show the assignment completeness of the four DESI samples, within unique tile groupings. One can observe that the patterns in the completeness correspond to the number of overlapping tiles. The black outline shows the edges of the area within which DESI dark or bright tiles have been defined, as of May 17th, 2024. The completeness maps can be compared to the values in \ref{['tab:statsvntile']}.
• Figure 3: Failure rate for each fiber on petal 5 for LRG (blue) and BGS (orange; labeled BGS_BRIGHT as that is the sample presented herein). Vetoed fibers, with a significantly higher failure rate, are colored dark red (LRG) or light red (BGS).
• Figure 4: The projected, completeness-corrected, density of each DESI LSS tracer in DR1 inside and outside of the imaging veto masks we apply. The plotted values are determined by counting the redshifts in each area and redshift bin and dividing by the fiber assignment completeness, the area in square degrees, and the width of the redshift bin. 'LS' denotes Legacy Survey, which defined a series of maskbit values. The difference between the dotted and solid curves illustrates the need to apply these masks, with the LRG case being the most extreme.
• Figure 5: In the top panel boxes show $\chi^2_{\rm null}$ values for the null test for BGS data, determined for each of the image property maps individually and the total across maps, in units of the mean $\chi^2_{\rm null}$ values for the same tests evaluated on the 25 null, i.e., uncontaminated mocks. Also shown are the totals obtained from the unweighted data. Results for the BASS/MzLS and DECam regions are shown separately in blue and red respectively. Dashed lines indicate the range of values obtained from the 25 mocks. The bottom panel shows the three cases where the null tests against imaging systematics are worst, in comparison to the expectation from mock analysis. The dashed curves display the results when no weights for imaging systematics were applied to the data, while the points with error bars display the cases where they were. The points thus display the residual trends in the final catalogs. The results plotted with a solid curve replace the SFD $E(B-V)$ map with the $E(B-V)_{\rm no CIB}$ when determining $\Delta$EBV GR.