Constraining primordial non-Gaussianity from DESI DR1 quasars and Planck PR4 CMB Lensing

TL;DR

This paper reports the first measurement of local-type primordial non-Gaussianity ($f_{\mathrm{NL}}$) from the cross-correlation of DESI DR1 spectroscopic quasars with Planck PR4 CMB lensing maps, using three tomographic redshift bins and a catalog-based angular power spectrum estimator. The authors model the PNG-induced scale-dependent bias $b(k,z)$ with a universality relation for $b_\Phi(z)$ and perform a non-Limber angular power spectrum calculation, including magnification and RSD contributions. They validate imaging systematics weights with extensive mocks, derive an optimal redshift weighting to boost $f_{\mathrm{NL}}$ sensitivity, and explore bias-model variations, finding $f_{\mathrm{NL}}=2^{+28}_{-34}$ for $p=1.6$ and $f_{\mathrm{NL}}=6^{+20}_{-24}$ for $p=1.0$, with a ~35% improvement over previous DR9 analyses. The results demonstrate DESI quasars’ power for inflationary physics and point to substantial gains with future DESI data releases, especially when combining cross- and auto-correlations while controlling systematics.

Abstract

We present the first measurement of local-type primordial non-Gaussianity from the cross-correlation between $1.2$ million spectroscopically confirmed quasars from the first data release (DR1) of the Dark Energy Spectroscopic Instrument (DESI) and the Planck PR4 CMB lensing reconstructions. The analysis is performed in three tomographic redshift bins covering $0.8 < z < 3.5$, covering a sky fraction of $\sim 20\%$. We adopt a catalog-based pseudo-$C_\ell$ estimator and apply linear imaging weights validated on noiseless mocks. Compared to previous analyses using photometric quasar samples, our results benefit from the high purity of the DESI spectroscopic sample, the reduced noise of PR4 lensing, and the absence of excess large-scale power in the spectroscopic quasar auto-correlation. Fitting simultaneously for the non-Gaussianity parameter $f_{\mathrm{NL}}$ and the linear bias amplitude in each redshift bin, we obtain $f_{\mathrm{NL}} = 2^{+28}_{-34}$ for a response parameter $p=1.6$, and $f_{\mathrm{NL}} = 6^{+20}_{-24}$ for $p=1.0$. These results improve the constraints on $f_{\mathrm{NL}}$ by $\sim 35\%$ compared to the previous analysis based on the Legacy Imaging Survey DR9. Our results demonstrate the statistical power of DESI quasars for probing inflationary physics, and highlight the promise of future DESI data releases.

Figures (18)

• Figure 1: Left: Total quasar-CMB lensing cross-correlation in the first tomographic bin ($0.8<z<2.1$). The fiducial model with $f_{\mathrm{NL}}=0$ is presented in blue, and with $f_{\textrm{NL}} = 50$ in magenta. Right: Contributions of each term to Eq. \ref{['eqn:total_model']} as a fraction of the fiducial model with $f_{\textrm{NL}} = 50$. Negative terms are shown as dashed lines.
• Figure 2: Data used for the cross-correlation (same as in Ref. de2025cosmology). Panel (a) shows the Planck PR4 lensing convergence map, $\kappa$, together with its smoothed mask, obtained by applying a Gaussian filter with $1^{\circ}$ FWHM. Panel (b) displays the DESI DR1 quasar number counts for the full sample spanning the redshift interval $0.8 \leq z \leq 3.5$, while the corresponding completeness mask is shown in panel (c). For visualization, we adopt a HEALPix resolution of $N_{\rm side}=128$, whereas all computations are performed at $N_{\rm side}=2048$. All maps are displayed in a Mollweide projection and are presented in the Galactic coordinate system.
• Figure 3: Normalized redshift distribution $n(z)$ of the DESI DR1 spectroscopic quasar sample. The different colors identify the three non-overlapping redshift bins.
• Figure 4: Monte Carlo normalization correction applied to the cross-spectra $C_\ell^{\kappa g_i}$ which amounts to an approximately constant amplitude change of $\approx 8\%$.
• Figure 5: Left: A zoom into the survey footprint. In darker blue, the DESI DR1 quasars, while in light blue are the galaxies making up one of the $500$ mocks. The agreement between the two footprint is very good, confirming the success of our procedure to replicate the DESI survey geometry. Right: Average angular power spectrum $C_\ell^{\kappa g}$ over $500$ mocks with the corresponding $1\sigma$ error bars. As desired, the mocks have a power spectrum that matches the input theory curve.