ACT Data Release 6: High-Resolution CMB Maps

• ACT DR6 is a comprehensive CMB dataset spanning 2017–2022, offering high-resolution maps and precise temperature, polarization, and lensing measurements.
• The release employs state-of-the-art instrumentation, advanced maximum-likelihood mapmaking, and robust simulation-based noise and covariance modeling.
• ACT DR6 enables stringent cosmological parameter estimation and inflationary tests through detailed lensing reconstructions and cross-survey data integration.

The Atacama Cosmology Telescope (ACT) Data Release 6 (DR6) constitutes the most comprehensive public dataset yet from ACT, spanning 2017–2022 and containing high-resolution, wide-sky maps and advanced derived products targeting cosmic microwave background (CMB) science. DR6 underpins new levels of precision in cosmological inference, enabling high signal-to-noise measurements of CMB temperature and polarization, lensing, power spectra, and derived cosmological parameters. The extensive sky coverage, advanced instrument performance, and robust methodological infrastructure implemented for DR6 are essential for the cross-validation of cosmological models and for the search of physics beyond the standard paradigm.

1. Instrumentation, Observations, and Mapmaking

ACT DR6 maps are generated from data taken with the Advanced ACT camera installed on the 6-meter off-axis Gregorian telescope at an altitude of 5190 m in Chile. The instrument comprises three refractive optics tubes (PA4, PA5, PA6), each utilizing dichroic, polarized AlMn Transition Edge Sensor (TES) bolometer arrays cooled to ~100 mK. Time-domain multiplexed SQUID readout supports a large pixel count and high-speed acquisition across three frequency bands: 98 GHz (f090), 150 GHz (f150), and 220 GHz (f220). Observational strategy features constant-elevation azimuthal scanning, providing partial sky coverage of 19,000 deg² (≈47%) between –60° < dec < +20°, with 10 µK·arcmin combined noise and arcminute (1.01′–2.07′ FWHM) resolution. Data selection discriminates between day and night periods, applying extensive quality cuts to mitigate atmospheric effects, thermal deformations, and detector anomalies.

The mapmaking pipeline applies a maximum-likelihood formulation in pixel space: $$\hat{a} = \left(A^T N^{-1} A + \Pi\right)^{-1} A^T N^{-1} d$$ where A encodes the pointing and systematic correction matrices, N is the time-domain noise covariance computed in frequency bins, and $\Pi$ regularizes high-contrast modes. Pointing is iteratively refined via planet calibrations and point-source fits, with residual jitter yielding minor beam inflation. Calibration integrates both relative (flat-field) and absolute (planet flux vs Planck) components to ensure consistent gains.

2. Noise Characterization and Map-Based Simulation

Noise in DR6 is nontrivial due to atmospheric correlations, scan strategy, and detector inhomogeneity. The principal features include large-scale red noise, frequency-frequency correlations (notably within dichroic arrays), spatial anisotropy ("stripiness"), and scale-dependent variance amid map-depth inhomogeneity. ACT DR6 incorporates advanced Gaussian map-based noise models employing tiles, isotropic wavelets, and directional wavelets, as implemented in the mnms simulation package (Atkins et al., 2023). These models use localized decomposition (Fourier or harmonic space) and empirically estimated block-diagonal covariance structures to reproduce observed noise profiles and their covariance over splits and frequencies: $$\mathcal{N}_{i,j}(\alpha,\beta) \approx \langle u_{i,j}^\alpha u_{i,j}^{\beta*} \rangle$$ Direct simulation ensembles generated by mnms are utilized for covariance estimation, analytic validation, lensing bias corrections, and robust power spectrum analysis.

3. Power Spectrum Estimation, Covariance Modeling, and Foregrounds

DR6 delivers multi-frequency TT, TE, EE, and BB bandpowers over ℓ ≈ 600–4500, adopting rigorous analytic and simulation-based covariance matrix prescriptions (Atkins et al., 10 Dec 2024). Critical advances include handling sharp features via the improved-narrow-kernel approximation (INKA), explicit accommodation of distinct signal and noise masks reflecting inhomogeneous depth, and simulation-based tuning of analytic covariances to the Monte Carlo realization spectra. The final corrected covariance matches Monte Carlo variance at the ≲1% level, advancing beyond prior homogeneous-depth approaches.

Foreground modeling is structured as a parametric, multi-component sum: $$D_\ell^{X_iY_j} = D_\ell^{\rm CMB,\,XY} + D_\ell^{\rm FG,\,X_iY_j}$$ with 14 baseline foreground parameters per band, describing thermal and kinematic Sunyaev–Zel'dovich effects (tSZ, kSZ), cosmic infrared background (CIB, both Poisson and clustered), radio point sources, Galactic dust, and SZ–CIB/radio cross-terms (Beringue et al., 6 Jun 2025). Model extensions span alternative SEDs, decorrelations, templates (AGORA/SPT), and CO lines. Robustness checks confirm cosmological parameters are stable under these variations (≤0.5σ for ΛCDM, ΛCDM + N_eff), while real-sky simulations from PySM/AGORA/Sehgal/BAHAMAS maintain integrity of foreground parameter estimation.

4. CMB Lensing: Methodologies, Cosmological Results, and Consistency

CMB lensing analysis applies a hybrid quadratic estimator [Hu & Okamoto 2003]. The reconstructed lensing potential $\phi_{LM}$ and convergence $\kappa_{LM}$ are obtained from minimum-variance coadds using cross-spectrum estimators that suppress noise bias. Foreground hardening via tSZ-profile nulling and ILC-based CIB mitigation, together with rigorous masking and MC bias subtraction, enable high-fidelity lensing bandpower estimation (Qu et al., 2023, Bashir et al., 22 Mar 2025).

ACT DR6 achieves a 43σ detection of CMB lensing over 9400 deg², measuring the lensing amplitude at the 2.3% level: $$A_{\rm lens} = 1.013 \pm 0.023 \quad \text{(relative to Planck 2018 ΛCDM)}$$ The derived growth parameter,

$$S_8^{\rm CMBL} = \sigma_8\, (\Omega_m/0.3)^{0.25} = 0.818 \pm 0.022$$

is in excellent agreement with Planck and ACT DR4 + WMAP analyses. Null tests include noise differences, frequency splitting, curl reconstruction, spatial and temporal splits, and bandpower–level consistency. Morphological analysis of the DR6 convergence map (Bashir et al., 22 Mar 2025) employing Minkowski functionals, contour Minkowski tensors, and Betti numbers demonstrates global consistency with ΛCDM, yet reveals persistent, scale-localized deviations—especially near the CMB Cold Spot. These anomalies do not coincide with Planck-identified regions and are suggestive of improved S/N and systematics control in DR6. Follow-up with large-scale structure surveys is warranted.

5. Inflationary Constraints and Theoretical Impact

DR6 sharpens bounds on inflationary parameters via high-precision spectral tilt and tensor-to-scalar ratio measurements, often in combination with Planck, BICEP/Keck, DESI BAO, and SPT (Ellis et al., 21 Oct 2025, Heidarian et al., 12 Jun 2025, Rehman et al., 21 Apr 2025). The observed preference for $n_s \approx 0.97$ (e.g., $n_s = 0.973 \pm 0.006$ in P-ACT-LB) challenges canonical α-attractor (Starobinsky/T-model) and supersymmetric hybrid inflation models; compatibility is restored for generalized α-attractors with higher monomial power (k ≥ 6), larger α (α ≤ 25 for E-model, α ≤ 11 for T-model), and quantum-gravity modifications (GUP parameter β ≳ 10¹³, leading to α ≤ 0.74 at β = 1.4 × 10¹³). Tensor-to-scalar ratio is constrained at $r_{0.05} < 0.036$ (95% CL), with minimal models predicting $r \lesssim 0.01$ (Rehman et al., 21 Apr 2025).

Minimal supersymmetric hybrid inflation parameters are tightly constrained: κ in 3×10⁻⁴–0.3, α_h = κ² N/(8π²) in 5×10⁻⁵–5×10⁻², and gauge-breaking scale M in (0.5–2) × 10¹⁵ GeV. Quantum-gravity corrections are increasingly motivated as ACT DR6 alone would otherwise exclude several canonical models at 68% CL.

6. Systematics, Consistency Tests, and Nested Sampling Analyses

ACT DR6 incorporates a blinding framework, extensive null tests, and Bayesian consistency evaluations against external datasets (Planck, BAO, DES-Y1, NPIPE) using nested sampling with PolyChord and the Cobaya likelihood interface (Ormondroyd et al., 2023). Principal cosmological parameters (σ₈, Ω_m, S₈) are consistent across prior choices (informative versus uniform), with shifts ≤0.3σ. Quantitative tension metrics (Bayes factor, suspiciousness S) exhibit p-values ≫5% in all dataset combinations, indicating no statistically significant tension. Nested sampling efficiently converges on posteriors, outperforming MCMC for broad priors and yielding robust evidence estimates.

Notable systematic is an internal ~0.2° discrepancy in PA5 f090/f150 polarization miscalibrations, partially absorbed by correlated priors but presently unexplained. In lensing, cross-spectrum, frequency-split, and curl null tests, systematic errors are subdominant to statistical uncertainties due to comprehensive simulation and model-bias correction pipelines.

7. Data Products, Accessibility, and Cross-survey Integration

ACT DR6 releases maps (I, Q, U), inverse-variance weightings, cross-linkage diagnostics, per-split depth maps, null test splits, component-separated CMB maps, and matched-filter products via LAMBDA, NERSC, interactive web atlases, and HiPS data sets. File formats include FITS, HDF5, and plain text, with explicit spatial index files and mask definitions. DR6 products overlap with Euclid, LSST, DESI, DES, HSC, SPT, enabling wide-ranging cross-correlation and multiwavelength science.

The validated power spectrum, lensing, and foreground model infrastructure, together with public simulation codes (mnms), position ACT DR6 as a robust backbone for high-resolution, wide-sky cosmology. The methodologies and pipeline components are directly transferable to next-generation experiments (Simons Observatory, CMB-S4), particularly regarding precision noise modeling, analytic–simulation covariance correction, and systematic control.

---

In summary, ACT Data Release 6 delivers advanced high-resolution CMB maps, power spectra, lensing reconstructions, and validated modeling infrastructure, enabling precise cosmological parameter estimation, robust consistency checks, and critical tests of inflationary and beyond-standard-model theories. The dataset represents a foundational resource and methodological template for ground-based CMB cosmology.