The Atacama Cosmology Telescope: A demonstration of CMB lensing measurement from daytime data

TL;DR

This work demonstrates a robust CMB lensing measurement using daytime Atacama Cosmology Telescope data from 2017–2022 (ACT DR6). Using a careful calibration strategy against nighttime data, extensive null tests, and a maximum-variance lensing estimator, the authors reconstruct the lensing signal with daytime observations, achieving a significance of $17\sigma$ in the range $40<L<763$ and obtaining $A_{lens}=1.045\pm0.063$. When combined with DESI BAO data, this yields $\sigma_8=0.826\pm0.027$, consistent with nighttime ACT results and Planck-based predictions. This daytime demonstration validates the inclusion of day data for future ACT analyses and motivates daytime-enabled approaches for next-generation surveys like the Simons Observatory.

Abstract

We present a cosmic microwave background (CMB) lensing power spectrum analysis using daytime data (11am-11pm UTC) gathered by the Atacama Cosmology Telescope (ACT) over the period 2017-2022 (ACT Data Release 6). This dataset is challenging to analyze because the Sun heats and deforms the telescope mirror, complicating the characterization of the telescope. We perform more than one hundred null and consistency checks to ensure the robustness of our measurement and its compatibility with nighttime observations. We detect the CMB lensing power spectrum at 17$σ$ significance, with an amplitude $A_\textrm{lens} = 1.045 \pm 0.063$ with respect to the prediction from the best-fit Planck-ACT CMB power spectrum $Λ$CDM cosmology. In combination with the Dark Energy Spectroscopic Instrument (DESI) Baryon Acoustic Oscillation (BAO) data, this corresponds to a constraint on the amplitude of matter fluctuations $σ_8 = 0.826 \pm 0.027$. The analysis presented here is especially relevant for ground-based millimeter-wave CMB experiments at the Atacama site, paving the way for future analyses making use of both nighttime and daytime data to place tight constraints on cosmological parameters.

Figures (8)

• Figure 1: Map of root-mean-square (RMS) noise per pixel for daydeep (top) and daywide (bottom) derived from inverse-variance maps (that describe the noise behavior on small scales Atkins_2023Naess_DR6_2025). Regions such as the left half of the southern daydeep patch are shallower due to data cuts prior to map-making, such as beam cuts (summarized in the beam-badness statistic; see Storer_thesisNaess_DR6_2025). The RMS maps, together with cross-linking and Galactic dust emission information, inform the choice of area that we model and simulate (encircled with gray contours). Further null tests inform the regions we use in our lensing power spectrum analysis: daywide South, shown in blue, with a mean depth of $24µK$-arcmin and effective sky fraction $f_\textrm{sky}=0.08$; and daydeep, shown in orange, with a mean depth of $8µK$-arcmin and $f_\textrm{sky}=0.02$.
• Figure 2: Maps from detector array-bands (PA) are combined using inverse-variance weights in harmonic space to form coadded daydeep and daywide sky maps, following the same approach as Frank_ACT_lensing_2024. This figure shows the weights $w_\ell$ applied to each array-band map as a function of multipole. They sum to unity at each multipole to preserve the CMB signal. For daydeep, the total weights assigned to each frequency are shown with solid lines. We shade in gray the CMB scales excluded in our analysis (for which we only include $600< \ell_{\textrm{CMB}}<3000$, informed by foreground contamination studies MacCrann_2024AbrilCabezas_2025). The PA6 daywide dataset is excluded from our analysis as it is too shallow.
• Figure 3: We validate the ACT DR6 daytime lensing power spectrum with a suite of null tests targeted at potential systematics in the daytime data. Top: The lensing power spectrum reconstructed from the difference between the daytime and nighttime maps. This is a stringent test: the signal is null within 1% of the lensing signal (shown in blue). Bottom: Difference between the lensing power spectrum reconstructed from the daytime data and the ACT dr6.02 nighttime-only lensing power spectra Kim_2025, in units of the nighttime-only bandpower errors $\sigma_L$. The PTEs for these null tests are calculated from the goodness-of-fit with respect to null, by Monte Carlo sampling from the covariance matrix within the analysis range $40<L<763$.
• Figure 4: Difference in the reconstructed lensing bandpowers between our baseline analysis ($600<\ell_\textrm{CMB}<3000$) and a reconstruction where only $\ell_\textrm{CMB}>1000$ modes in polarization are used, divided by the error in the baseline bandpowers $\sigma_L$. This test is motivated by the fact that the daytime beams are approximate because the polarization beam includes a transfer function at $\ell_\textrm{CMB} < 1000$ that should only be present in temperature. We find consistency with null, with a PTE of 0.41. Excluding $\ell_\textrm{CMB}<1000$ modes in polarization minimally affects the sensitivity of our measurement (SNR of 16$\sigma$, c.f. baseline is 17$\sigma$).
• Figure 5: Histogram of PTE values for all our null tests. The top panel shows the null tests that refer to the measured gradient component of the lensing deflection field, while the bottom panel corresponds to the curl component. Both distributions are consistent with a uniform distribution (Kolmogorov--Smirnov test PTE of 95% and 9%, respectively; with the caveat that this test ignores correlations between the different tests). None of our null tests lie outside the range $0.001<\textrm{PTE}<0.999$.