First Detection of the BAO Signal from Early DESI Data

TL;DR

The paper reports the first high-significance BAO detection from the DESI-M2 early data, using LRG and BGS samples to validate the DESI spectrograph and data-management pipeline. It adopts a robust analysis chain—two-point clustering with the Landy–Szalay estimator, density-field reconstruction, and template-based BAO fitting with an isotropic scale parameter $\alpha$—and validates covariance modeling with RascalC against mock catalogs. The LRG sample yields ~5$\\sigma$ BAO detection at ~1.7% precision, while the BGS sample yields ~2.3–3.0$\\sigma$ at ~2.6% precision, with reconstruction strengthening the BGS result more noticeably. These early results demonstrate DESI’s exceptional statistical power, provide end-to-end quality assurance for the data-management pipeline, and forecast sub-percent BAO precision (≈0.29%) for the Year 5 data when the full survey is analyzed.

Abstract

We present the first detection of the baryon acoustic oscillations (BAO) signal obtained using unblinded data collected during the initial two months of operations of the Stage-IV ground-based Dark Energy Spectroscopic Instrument (DESI). From a selected sample of 261,291 Luminous Red Galaxies spanning the redshift interval 0.4 < z < 1.1 and covering 1651 square degrees with a 57.9% completeness level, we report a ~5 sigma level BAO detection and the measurement of the BAO location at a precision of 1.7%. Using a Bright Galaxy Sample of 109,523 galaxies in the redshift range 0.1 < z < 0.5, over 3677 square degrees with a 50.0% completeness, we also detect the BAO feature at ~3 sigma significance with a 2.6% precision. These first BAO measurements represent an important milestone, acting as a quality control on the optimal performance of the complex robotically-actuated, fiber-fed DESI spectrograph, as well as an early validation of the DESI spectroscopic pipeline and data management system. Based on these first promising results, we forecast that DESI is on target to achieve a high-significance BAO detection at sub-percent precision with the completed 5-year survey data, meeting the top-level science requirements on BAO measurements. This exquisite level of precision will set new standards in cosmology and confirm DESI as the most competitive BAO experiment for the remainder of this decade.

Figures (10)

• Figure 1: Redshift distribution of the four primary DESI tracers, from the DESI-M2 clustering catalogs.
• Figure 2: Footprints of the DESI-M2 BGS Bright (top) and LRG (bottom) clustering samples, color-coded by completeness weights. The total areas highlighted by the pink color represent the final DESI Y5 expected footprint. The specific DESI-M2 areas covered by the BGS Bright and LRG samples are respectively 3677 deg$^2$ and 1651 deg$^2$.
• Figure 3: [Top] Map of a single LRG EZmock realization with the pixel probability of the HEALPix mask. [Bottom] Dispersion of $1000$ LRG EZmocks, after application of the different masks described in Section \ref{['subsec:masks']}, compared with the actual LRG DESI-M2 redshift distribution.
• Figure 4: Monopole of the LRG two-point correlation function before reconstruction, as measured from the DESI-M2 sample (blue dots) and from the average of 1000 EZmocks (orange dots). Data errorbars are obtained from a jackknife covariance directly inferred from DESI-M2 LRGs, while mock errorbars are drawn from the LRG EZmock sample covariance. As mentioned in the main text (Section \ref{['subsec:calibration']}), the $\sim 10\%$ difference near the $\sim 20~\,h^{-1}{\rm Mpc}$ peak is not surprising, as these mocks were tuned with an earlier version of the DESI data. Hence, in the present work mocks are only used for validation purposes.
• Figure 5: Two-point correlation function measurements of the four DESI tracers, obtained from the DESI-M2 sample. Errorbars are derived from the diagonal of the corresponding covariance matrixes, although we caution the reader of a significant bin-to-bin correlation in these measurements. Model curves are simple damped linear theory predictions that indicate the expected overall clustering amplitude and BAO damping typical at the mean redshift of the target samples.