SPT-3G D1: High-Fidelity CMB & Cluster Data

• SPT-3G D1 dataset is a public release of deep, multi-frequency CMB observations covering ~4% of the sky with unprecedented sensitivity in TT/TE/EE spectra.
• It employs a rigorously validated pipeline with precise calibration via Planck cross-spectra, detailed beam measurements, and extensive null tests to ensure data accuracy.
• The dataset underpins advanced cosmological and cluster science, yielding tight constraints on parameters like H0 and σ8 and enabling joint analyses with BAO and multiwavelength surveys.

The SPT-3G D1 dataset is a public release of high-fidelity cosmic microwave background (CMB) observations from the third-generation polarization-sensitive camera (SPT-3G) on the South Pole Telescope (SPT). Covering ~4% of the sky with uniform, deep mapping, SPT-3G D1 delivers the most sensitive high-ℓ measurements to date of CMB temperature and polarization (TT/TE/EE) power spectra, and provides a rigorously validated pipeline for cosmological inference, cluster science, and joint ground-based CMB analyses (Camphuis et al., 25 Jun 2025, Posta et al., 2021, Kornoelje et al., 21 Mar 2025).

1. Survey Design and Instrumentation

SPT-3G D1 comprises two full austral-winter seasons (2019–2020) of observations targeting the 38° × 38° Main field (1,500 deg²) centered at RA = 23h 30m, Dec = –55°. The scan protocol consists of constant-elevation sweeps with daily azimuthal shifts, optimizing for azimuthal cross-linking and uniform coverage (~60% of available SPT observing time) (Camphuis et al., 25 Jun 2025). SPT-3G employs ~16,000 polarization-sensitive transition-edge sensors (TES) distributed among 68 dual-polarization modules. All observations are simultaneous in three frequency bands: 95 GHz (FWHM ~1.6′), 150 GHz (~1.2′), and 220 GHz (~1.0′). The bandpasses provide approximate fractional bandwidths of 5%, 8%, and 15% (Main field) or 25% (Deep field), respectively. The effective map depths reach 3.3 μK·arcmin (T) and 5.1 μK·arcmin (Q/U), the deepest for any high-multipole CMB TT/TE/EE dataset to date (Camphuis et al., 25 Jun 2025, Kornoelje et al., 21 Mar 2025).

2. Data Processing, Mapmaking, and Calibration

The SPT-3G D1 pipeline calibrates the primary map gain at 150 GHz via cross-spectra with Planck PR3 at 143 GHz, yielding 0.2% calibration uncertainty. Gains for 95 GHz and 220 GHz are transferred via internal calibrators and atmospheric opacity monitoring (0.3% and 0.5% uncertainty, respectively). Beam transfer functions $b_\ell$ are precisely measured through planet observations and characterized in orthogonal eigenmodes (five per band), with uncertainties <0.5% up to $\ell\approx 3000$ (Camphuis et al., 25 Jun 2025). Time-ordered data (TOD) undergo electronic response correction, low-frequency drift removal (first-order polynomial per scan), and atmospheric common-mode subtraction through principal component analysis. A 20 Hz low-pass filter suppresses high-frequency noise. The cleaned TOD are binned with inverse-variance weighting onto HEALPix maps (Nside = 4096; Main field) or rectilinear CAR pixels (D1 Deep field), yielding T, Q, and U products for each frequency and jackknife split (Posta et al., 2021, Kornoelje et al., 21 Mar 2025).

Bright point sources (>6 mJy at 150 GHz) are masked using 5′ radius discs with inpainting performed via the CORK algorithm, matching the local covariance. Validation comprises >100 null tests (time, detector, scan-direction splits) on TT, TE, and EE bandpowers, with unblinding contingent on the absence of significant residuals at $\ell>400$ (Camphuis et al., 25 Jun 2025).

3. Power Spectrum and Cluster Catalog Construction

Power spectrum estimation utilizes a curved-sky pseudo–$C_\ell$ ("MASTER") approach. Binned spectra $\hat S_b$ are computed for TT ($400\leq\ell\leq 3000$), TE, and EE ($400\leq\ell\leq 4000$) in bins optimized for tracking acoustic peaks, with noise bias $N_\ell^{XY}$ estimated from jackknife differences. Mode-coupling and transfer functions are applied to yield unbiased bandpowers: $$C_b^{XY} = \sum_{b'} (M^{-1})_{bb'} \left[\sum_{\ell} B_{b'\ell} (\hat{\tilde C}_\ell^{XY}-N_\ell^{XY})\right].$$ Lensing corrections convolve the unlensed theory spectra with the lensing potential $C_{\phi\phi}$ within the likelihood analysis (Camphuis et al., 25 Jun 2025).

In the D1 Deep field, clusters are detected via minimum-variance matched filters exploiting the thermal Sunyaev-Zel'dovich effect, with detection significance $\xi$ and mass-observable scaling ($\zeta$–$M_{500c}$) calibrated against optical/IR redshifts and confirmed using ancillary datasets (DES Y6, Spitzer/IRAC). Completeness and purity are quantified with mock maps, and extensive dust contamination mitigation is achieved by leveraging the 220 GHz channel and SPIRE-based constrained internal linear combinations (Kornoelje et al., 21 Mar 2025).

4. Covariance, Systematics, and Data Products

Covariances are constructed semi-analytically by summing Gaussian sample variance (from ΛCDM best fit), instrumental noise, mode-coupling, beam eigenmode, and calibration uncertainties. Independent validation with 2,000 end-to-end simulations yields diagonal accuracy better than 1% and off-diagonal elements to within 5% (Camphuis et al., 25 Jun 2025). Systematics tracked in the total error budget include beam eigenmode uncertainties (≤0.5% for TT to $\ell\approx2000$, ≤1% at higher multipoles; 0.3% for TE/EE), absolute calibration, polarization-angle error (≤0.3° per detector set), temperature-to-polarization leakage (<0.5 μK rms in Q/U), and filter function uncertainties (0.5% from simulations, marginalized in final likelihoods).

Released data products encompass ASCII bandpower files for TT, TE, EE (with $\ell$ bins, D_b, $\sigma_b$), full covariance matrices, window functions $B_{b\ell}$, beam transfer function FITS tables for all bands, point-source and Galactic masks (HEALPix and CAR), filter transfer functions, Python notebooks demonstrating typical analysis steps, and matched-filter cluster catalogs (minimum-variance and dust-nulled) (Camphuis et al., 25 Jun 2025, Kornoelje et al., 21 Mar 2025). All files adhere to reproducibility and software convention standards (Python ≥3.9, numpy, healpy, jax, candl, SPTlite, GetDist).

5. Cosmological, Astrophysical, and Cluster Science Applications

SPT-3G D1 achieves the deepest high-ℓ CMB TT/TE/EE measurements over a wide field, enabling stringent ΛCDM and extended cosmological model constraints. For the base model, SPT-3G alone yields $H_0=66.66\pm0.60$ km/s/Mpc, $6.2\sigma$ away from local SH0ES values. When combining SPT, ACT-DR6, and Planck, the constraint tightens to $H_0=67.24\pm0.35$ km/s/Mpc; cluster amplitude $\sigma_8=0.8137\pm0.0038$ (Camphuis et al., 25 Jun 2025). The dataset supports direct likelihood integration into MCMC cosmological inference, with recommended use of the "candl" JAX-based framework. Joint analyses with baryon acoustic oscillation (BAO) data (DESI-DR2) probe deviations from ΛCDM at the 2–3$\sigma$ level in geometric curvature, lensing amplitude, and the dark energy equation of state, and indicate mild model preference for modified early-universe or recombination physics (Camphuis et al., 25 Jun 2025, Posta et al., 2021).

The cluster catalogs derived from SPT-3G D1 (in concert with SPTpol data) provide the highest redshift and lowest mass median Sunyaev-Zel'dovich-selected cluster sample to date, with extensive validation against deep optical/IR surveys and accurate dust bias control due to multi-frequency coverage (Kornoelje et al., 21 Mar 2025).

6. Access, Usage Recommendations, and Integration with Other Surveys

The full SPT-3G D1 dataset can be obtained via the SPT data portal or Zenodo DOI. Recommended workflow mandates downloading the bandpower, covariance, window, beam, and mask files, validation with supplied Python scripts, and convolution of theoretical $C_\ell$ (including lensing) with instrument response before likelihood evaluation. Users are advised to include calibration and beam-error nuisance parameters as described in the README. For joint cosmological analysis, SPT-3G D1 likelihoods integrate directly with CosmoMC or MontePython frameworks, with cross-validation achieving mutual ΛCDM parameter consistency to ≤0.2$\sigma$ between SPT, Planck, and ACT likelihoods (Camphuis et al., 25 Jun 2025).

For SZ-selected cluster science, FITS temperature maps, weight/mask files, and associated catalogs support confirmation and cross-matching with large-scale structure and multiwavelength surveys. The dataset's guides and example codes facilitate reproducibility and methodological rigor for precision cosmology and astrophysics (Kornoelje et al., 21 Mar 2025).

7. Relation to Prior SPT Datasets and Scientific Impact

The Main field SPT-3G D1 release supersedes all previous SPT TT/TE/EE datasets in sensitivity and precision at small angular scales, and represents an advance over earlier "D1" power spectrum releases (e.g., SPT-3G half-season polarization, 2018) used in extensions to ΛCDM such as early dark energy constraints (Posta et al., 2021). Integration with ACT DR6 and Planck places ground-based CMB science at parity with satellite measurements for select cosmological parameters. The D1 dataset thus underpins a new era of high-multipole, ground-based, and multi-probe CMB cosmology, cluster science, and cross-correlation with galaxy and lensing surveys (Camphuis et al., 25 Jun 2025, Kornoelje et al., 21 Mar 2025).