SPT-3G D1: Compton-$y$ maps using data from the SPT-3G and Planck surveys

TL;DR

This work constructs high-resolution Compton-$y$ maps by optimally combining SPT-3G D1 data with Planck all-sky data using a 2D harmonic-space linear combination that preserves the tSZ signal while suppressing CMB and foregrounds. The authors implement minimum-variance, CMB-deprojected, and CIB-deprojected variants within a constrained ILC framework, validate them with auto- and cross-spectra, cross-correlations with unWISE galaxies, and cluster-stack analyses, and provide split maps to mitigate noise bias. They demonstrate that the CMB- and CIB-deprojected maps offer reduced foreground contamination for cross-correlations and large-scale structure studies, while the MV map remains optimal for small-scale, low-foreground analyses such as cluster pressure profiling. The resulting data products enable precise power-spectrum measurements, cross-correlations, and detailed studies of hot gas and baryon thermodynamics, with clear caveats on CIB modeling and plans for future expansion with upcoming facilities like the Simons Observatory.

Abstract

We present thermal Sunyaev-Zel'dovich (tSZ) Compton-$y$ parameter maps constructed from two years (2019-2020) of observations with the South Pole Telescope (SPT) third-generation camera, SPT-3G, combined with data from the Planck satellite. Using a linear combination (LC) pipeline, we obtain a suite of reconstructions that explore different trade-offs between statistical sensitivity and suppression of astrophysical contaminants, including minimum-variance, CMB-deprojected, and CIB-deprojected $y$-maps. We validate these maps through different statistical techniques such as auto- and cross-power spectra with large-scale structure tracers as well as stacking on cluster locations. These tests are used to understand the balance between noise and astrophysical foreground residuals (such as the CIB) in combination with the recovery of the tSZ signal for different maps. For example, results from stacking at the location of clusters confirm the robustness of the recovered tSZ signal over the $\sim 1500\: {\rm deg}^2$ SPT-3G survey field used in this analysis. The high-resolution and low-noise maps produced here provide an important cosmological tool for future studies, including measurements of the Compton-$y$ map power spectrum, cross-correlations with other tracers of the large-scale structure, detailed modeling of cluster pressure profiles, and study of the thermodynamic state of the baryons in the Universe.

Figures (8)

• Figure 1: Beam-deconvolved 1D noise power spectra for temperature data for different Planck (solid lines) and SPT (dashed lines) frequency channels. The Planck noise curves are from within the SPT-3G patch on the sky. As pointed out in Sec. \ref{['subsub:sptmapmaking']}, due to the high-pass filtering step during our map-making, SPT maps lose modes below $\ell < 300$. The black curve shows the expected CMB power spectrum using the 2018 best-fit Planck cosmology for comparison.
• Figure 2: The Planck and SPT-3G combined minimum-variance (MV) version of the Compton-$y$ map provided with this work. We can see numerous galaxy clusters prominently visible as reddish features, resulting from the tSZ effect.
• Figure 3: Auto-power spectra of different Compton-$y$ maps constructed in this work (left panel). The $x$-axis has been slightly offset in the left panel to aid the eye in comparing different curves. Cross-power spectra between combinations of the Compton-$y$ maps (right panel). Note that all the curves lie above the theory curve from Agora simulations because the maps are noise- and foreground-dominated at most scales (see Sec. \ref{['subsub:cibcross']}). Here $D_\ell = \frac{\ell (\ell + 1)}{2\pi} C_\ell$. Shaded bands represent the Gaussian error bars on the power spectra. The black solid curve is the Compton-$y$ power spectrum from Agora simulations. Dashed curves are the corresponding split auto- and cross-spectra with two data splits to remove the noise bias and isolate the signal and residual foreground components. We also show measurements of the Compton-$y$ power spectrum from Efstathiou2025Raghunathan2026 as purple and grey points respectively for comparison.
• Figure 4: Cross-power spectra between our MV (blue) and CIB-deprojected Compton-$y$ maps (green, purple, and red) and 545 GHz CIB maps from Lenz2019. Different colors for the CIB-deprojected $y$ maps correspond to different dust emissivity index $\beta_d$ considered for the CIB while deprojecting it in our LC algorithm. The error bars are computed within the Gaussian approximation. While we see a positive correlation for the MV $y$ map, correlation with the CIB-deprojected maps is smaller and can vary based on the $\beta_d$ value. While this points to potentially low level of CIB residuals in our CIB-deprojected Compton-$y$ maps, the differences in the residuals show the complex nature of the CIB and difficulties in modeling it accurately.
• Figure 5: Cross-correlation between the Compton-$y$ maps from this work and the unWISE blue and red galaxy samples in dashed and solid lines respectively. The error bars are calculated within the Gaussian approximation. The red sample peaks at higher redshift ($z \sim 1$) than the blue sample which predominantly traces galaxies at relatively low redshift ($z \sim 0.6$). Different colors show results for the MV, CMB-deprojected, and CIB-deprojected $y$-maps. The lower correlation observed for the CIB-deprojected maps than MV and CMB-deprojected maps, especially for the red galaxy sample at higher redshift where CIB contributes strongly, is consistent with suppressed CIB residuals in the former maps. This result is a preview of an upcoming analysis from Maniyar et al. (in preparation).