Estimation and mitigation of foregrounds in projected kSZ velocity reconstruction

Abstract

The kSZ effect has recently emerged as a powerful probe for precision cosmology through its ability to reconstruct the large-scale velocity field. In particular, the kSZ-reconstructed velocity-galaxy cross-correlation is sensitive to signatures of primordial non-Gaussianity through its imprint on the galaxy bias. The kSZ velocity reconstruction is performed using small-scale information from CMB temperature and galaxy overdensities. As the sensitivity of these measurements improves, systematic effects such as extragalactic foreground contamination present in CMB maps become increasingly important. We present a study of foreground biases to the kSZ-reconstructed velocity-galaxy cross-correlation. We derive the relevant foreground contributions from the thermal Sunyaev-Zel'dovich effect and the cosmic infrared background, modeling them using a halo model description of the dominant one- and two-halo terms. We compare our analytic predictions to measurements obtained using ACT DR6 temperature maps and DESI Legacy Imaging Survey galaxies, finding qualitative agreement. We introduce a parity-odd estimator constructed from antisymmetric combinations of tomographic velocity-galaxy correlations and show analytically that, under the Limber approximation, this estimator entirely cancels the foreground contamination while preserving the full cosmological signal without loss of signal-to-noise. Finally, we apply this parity-odd estimator to the data combination mentioned above and show that the fit to the velocity-galaxy correlation is dramatically improved compared to the analysis without mitigation; our estimator detects the signal at 11$σ$, with an amplitude consistent with recent studies.

Figures (9)

• Figure 1: Analytic estimates of the foreground bias to the cross-correlation of kSZ-reconstructed velocity and galaxies for a redshift bin centered at $\bar{z}=0.42$, corresponding to the first bin of the 16-bin case in Ref. McCarthy_kSZ2_25. We show the contributions from the 1-halo (dashed) and 2-halo (dotted) terms for tSZ (left), CIB (center), and their sum (right). The velocity field has been reconstructed using modes up to $\ell_{\text{max}} = 8000$ at frequencies of 150 GHz (orange) and 90 GHz (purple). The mass cut used in these computations is $M_{\text{vir}} = 3\times 10^{14}M_\odot$. We remind the reader that the "2-halo" contribution quantifies correlations between one large-scale galaxy leg in one halo, and two small-scale foreground and galaxy legs in another halo; this is clearly dominant, as expected.
• Figure 2: Redshift dependence of the foreground biases to the cross-correlation of kSZ-reconstructed velocity and galaxies at an observing frequency of $150\,\mathrm{GHz}$. The panels show the contributions from tSZ (left), CIB (center), and their sum (right), for tomographic bins centered at $z=[0.4, 0.8, 1.2, 1.7,2.2]$ (dark blue through yellow) with equal widths in comoving distance $\Delta \chi = 300\,\mathrm{Mpc}$. Calculations are done using a mock galaxy tracer with a constant minimum stellar mass threshold of $M_{\text{min}}^* = 10^{11.5}M_\odot$ across redshifts. The velocity reconstruction uses modes up to $\ell_{\text{max}} = 8000$, and the electron pressure profile is modeled using the Battaglia prescription.
• Figure 3: Comparison between our analytic predictions and measurements obtained using ACT DR6 CMB temperature maps and galaxies from DESILS, processed with the same pipeline as in Ref. McCarthy_kSZ2_25, at a frequency of $150\,\mathrm{GHz}$. The predictions shown are not fits to the data, instead we use two parametrizations of the electron pressure profile, the "Amodeo" (purple lines) and "Battaglia" (orange lines) models introduced in Sec. \ref{['sec:tsz_modeling']}, and the DESILS-like HOD and CIB models presented in Secs \ref{['sec:galaxy_HOD']} and \ref{['sec:cib_modeling']}, respectively. We show separately the tSZ (dashed) and CIB (dotted) contributions as well as their sum (solid). The comparison is performed for the same-redshift-bin configuration across the 16 tomographic bins used in Ref. McCarthy_kSZ2_25. The analytic calculations use a higher mass cut of $M_{\text{vir}} = 3 \times 10^{14}\,M_\odot$, chosen to mimic the cluster masking procedure applied to the CMB maps, and the velocity reconstruction uses modes up to $\ell_{\text{max}} = 8000$.
• Figure 4: Comparison to data of our analytic calculations using CMB data from ACT DR6 at a single frequency of 90 GHz and using DESILS as the galaxies. We show our predictions using the same configuration as Fig. \ref{['fig:data_150']} but computed at $90\,\text{GHz}$.
• Figure 5: Ratio between the predicted foreground contribution and the expected kSZ velocity–galaxy cross-correlation signal for the neighboring-bin configuration, where the velocity reconstruction is performed in redshift bin $\alpha$ and the galaxy density tracer is in bin $\beta$. All predictions are shown for an observing frequency of $150\,\mathrm{GHz}$. The foreground calculations assume a mass cut of $M_{\text{cut}} = 3 \times 10^{14}\,M_\odot$ to mimic the cluster-masking procedure and use modes up to $\ell_{\text{max}} = 8000$ for the velocity reconstruction. The theoretical signal has been computed using a modified version of ReCCO with $f_{\text{NL} }= 0$. We show that, across the scales where the signal-to-noise is concentrated, the foreground contamination can reach the amplitude of the predicted signal.