SPT-3G D1 2019-2020 CMB Data Release

• SPT-3G D1 is a groundbreaking data release offering precise ground-based CMB measurements that tighten constraints on neutrino masses and the Hubble constant.
• Advanced analysis techniques, including pseudo-Cℓ estimation and Fourier de-projection, ensure high-fidelity extraction of TT, TE, EE, and lensing spectra.
• The release reduces uncertainties by up to 35%, setting new benchmarks for cosmological parameter estimation in ground-based CMB surveys.

The SPT-3G 2019-2020 (D1) data release encompasses the most precise ground-based measurements to date of the cosmic microwave background (CMB) temperature (TT), E-mode polarization (EE), temperature-polarization cross spectrum (TE), and CMB lensing-potential ($ϕϕ$) power spectra. Derived from two years of observations (2019–2020) of the 1500 deg$^2$ SPT-3G Main field, the D1 release significantly tightens constraints on fundamental cosmological parameters, notably the sum of neutrino masses ($\sum m_ν$), Hubble constant ($H_0$), and large-scale structure amplitude ($σ_8$), reaching or surpassing the statistical power of space-based Planck and other advanced ground-based surveys (Camphuis et al., 25 Jun 2025, Gorbunov et al., 22 Jan 2026).

1. Observational Strategy and Data Acquisition

The SPT-3G D1 observations utilized the South Pole Telescope’s third-generation camera (SPT-3G) targeting a contiguous 4% fraction of the sky centered on the South Pole. The TT, TE, EE, and $\phi\phi$ power spectra were constructed from maps with $0.5'$ pixel resolution, following stringent protocols for noise mitigation and systematics control.

• Raw Data Processing: Detector timestreams were deconvolved for time constants and filtered at $<$25 Hz, then projected via HEALPix tessellation.
• Calibration: Daily calibration cycles used an internal thermal source, with absolute scaling derived from cross-power comparisons with Planck PR3 TT over $600<\ell<1800$.
• Beam Characterization: Composite beam profiles was determined with observations of Mars and Jupiter; final transfer function $B_\ell$ is symmetrized in harmonic space and corrected for $^2$00.5% cross-polar leakage.
• Sky Coverage: The full analysis utilizes a $^2$11500 deg$^2$2 area, the largest deeply mapped CMB region in a ground-based TT/TE/EE survey to date.

2. Data Analysis Pipeline and Systematics Suppression

The D1 data release is distinguished by its updated MUSE map-making pipeline, designed to optimize EE-based lensing reconstruction and expand multipole reach. The pipeline executes:

• Point Source and Ground Pickup Mitigation: Bright sources (S/N$^2$35) are masked and inpainted while atmospheric and ground-synchronous modes are excised via polynomial detrending and singular value decomposition. A Fourier-based “scan-synchronous template” de-projection is implemented for improved ground pick-up suppression.
• Pseudo-$^2$4 Estimation: Final TT, TE, EE power spectra are synthesized from six auto- and cross-frequency combinations (95, 150, 220 GHz) using MASTER estimators, with temperature-to-polarization leakage corrected by simulation-derived transfer functions.
• Noise Modelling and Covariance Construction: Noise spectra utilize half-mission split-map jackknifes. Covariance matrices incorporate sample variance, instrument noise, connected trispectrum (“$^2$5”) terms according to Camphuis et al. (2022), and are validated against 1000 Monte Carlo simulations to within 5% precision (Camphuis et al., 25 Jun 2025).

3. Power Spectrum and Bandpower Measurements

SPT-3G D1 provides angular power spectra in $^2$6 convention ($^2$7) covering expanded multipole ranges:

• TT: $^2$8 (30 bins)
• TE, EE: $^2$9 (40 bins each)
• Lensing MV ($\sum m_ν$0): $\sum m_ν$1 (10 bins)

Representative bandpower values and uncertainties from the D1 release:

$\sum m_ν$2	$\sum m_ν$3 ($\sum m_ν$4K$\sum m_ν$5)	1$\sum m_ν$6 uncertainty
500	5400	80
...	...	...
1750	3100	55

$\sum m_ν$7	$\sum m_ν$8 ($\sum m_ν$9K$H_0$0)	1$H_0$1 uncertainty
500	160	8
...	...	...
1750	90	8

$H_0$2	$H_0$3 ($H_0$4K$H_0$5)	1$H_0$6 uncertainty
1800	140	6
...	...	...
3300	60	10

$H_0$7	$H_0$8	1$H_0$9 uncertainty
100	2.1	0.30
...	...	...
1600	0.5	0.12

The D1 release achieves $σ_8$010 $σ_8$1K$σ_8$2 bin errors in EE at $σ_8$3, surpassing prior SPT-3G and ACT DR6 results particularly in high-$σ_8$4 regimes.

4. Covariance Structure and Likelihood Implementation

The joint analysis exploits a full data vector

$σ_8$5

with Gaussian covariance

$σ_8$6

Block diagonalization is performed to treat bandpowers from distinct experiments as independent; off-diagonal correlations in D1 are $σ_8$7 within spectra and $σ_8$8 between lensing and primary spectra.

Cosmological inference relies on the standard multivariate Gaussian bandpower likelihood:

$σ_8$9

Bandpower model predictions $\phi\phi$0 are generated with the CLASS Boltzmann solver, with parameter space spanning $\phi\phi$1.

5. Cosmological Constraints and Comparative Interpretation

The D1 spectra, in combination with multi-probe datasets (Planck PR3/PR4, DESI DR2, DES Y1, Pantheon+), yield refined cosmological parameters (Camphuis et al., 25 Jun 2025, Gorbunov et al., 22 Jan 2026):

Parameter	SPT-3G D1 TT/TE/EE+$\phi\phi$2 (best fit)
$\phi\phi$3	$\phi\phi$4
$\phi\phi$5	$\phi\phi$6
$\phi\phi$7	$\phi\phi$8
$\phi\phi$9	$0.5'$0
$0.5'$1	$0.5'$2
$0.5'$3	$0.5'$4
$0.5'$5 (km/s/Mpc)	$0.5'$6
$0.5'$7	$0.5'$8

• Neutrino Mass Constraint: Replacing the 2018 spectra with the D1 release tightens the upper bound on $0.5'$9 to $<$0 eV (95% CL) when jointly analyzed with DESI DR2 BAO and other probes; the posterior peaks at zero [(Gorbunov et al., 22 Jan 2026), Table 3a].
• Parameter Shifts: D1 achieves $<$125–35% reduction in uncertainty for $<$2, $<$3, and $<$4 compared to previous releases. The inferred $<$5 and $<$6 values are shifted upwards by %%%%77$^2$78%%%%, driving the neutrino-mass posterior toward the lower prior bound.

Comparison with Planck PR3/PR4 and ACT DR6 demonstrates agreement within $<$92%; D1’s precision matches or exceeds prior ground-and space-based constraints on $600<\ell<1800$0 and $600<\ell<1800$1. Notably, residual $600<\ell<1800$2 tension between CMB and BAO results (DESI-DR2) persists in baseline $600<\ell<1800$3CDM and is modestly relaxed in extended models.

6. Validation and Null Tests

The integrity of the D1 measurements is verified via extensive “blind” null and systematics tests:

• Detector and Time Splits: TT/TE/EE spectra are consistent with zero when split across detector polarization arrays and season halves (variations $600<\ell<1800$40.2$600<\ell<1800$5 per bin).
• Scan Direction and Foreground Checks: Ground-pickup residuals are sub-1% of signal; frequency cross-jackknife differences show no foreground leakage.
• End-to-End Simulations: 1000 realizations accurately recover input $600<\ell<1800$6CDM parameters ($600<\ell<1800$70.1$600<\ell<1800$8 deviations).
• Post-unblinding Corrections: Minor quadrupole leakage ($600<\ell<1800$90.3 $B_\ell$0K) and transfer-function corrections at $B_\ell$1 identified and remedied, shifting best-fit cosmology by $B_\ell$20.1$B_\ell$3.

These tests, combined with blind analysis protocols, ensure robust inference of astrophysical and cosmological signals.

7. Significance and Implications

The SPT-3G D1 2019-2020 release establishes new benchmarks in ground-based CMB analysis:

• Multipole Reach and Sensitivity: First sub-10 $B_\ell$4K$B_\ell$5 errors in EE and most precise TE spectrum at high-$B_\ell$6; expanded TT/TE/EE coverage to $B_\ell$7.
• Cosmological Parameter Power: Ground-based CMB-only constraints match space-based Planck for $B_\ell$8 and $B_\ell$9; combined SPT+ACT+Planck yields $^2$00 km/s/Mpc.
• Neutrino Mass Sensitivity: The D1 analysis is the first to push the $^2$01 posterior mode to zero with tightest upper bounds from CMB+BAO to date, indicating preference for quasi-negative neutrino masses in the SPT analysis context.
• Pipeline Advances: Adoption of MUSE map-making, advanced foreground mitigation, and expanded cross-experiment joint likelihood construction represent methodological progress.

A plausible implication is that systematic differences between SPT-3G 2018 and D1 reductions (distinct map-making and filtering approaches) play a nontrivial role in driving the shift in neutrino-mass constraints and parameter posteriors. This underscores the necessity of cross-validating analysis pipelines for next-generation CMB experiments. The foundation established by SPT-3G D1 is expected to directly inform future cosmological analyses leveraging increased sensitivity and sky coverage (Camphuis et al., 25 Jun 2025, Gorbunov et al., 22 Jan 2026).