The Atacama Cosmology Telescope: Detection of the pairwise kinematic Sunyaev-Zel’dovich effect with SDSS DR15 galaxies

We present a $5.4\sigma$ detection of the pairwise kinematic Sunyaev-Zeldovich (kSZ) effect using Atacama Cosmology Telescope (ACT) and Planck CMB observations in combination with Luminous Red Galaxy samples from the Sloan Digital Sky Survey (SDSS) DR15 catalog. Results are obtained using three ACT CMB maps: co-added 150 and 98 GHz maps, combining observations from 2008–2018 (ACT DR5), which overlap with SDSS DR15 over 3,700 sq. deg., and a component-separated map using night-time only observations from 2014–2015 (ACT DR4), overlapping with SDSS DR15 over 2,089 sq. deg. Comparisons of the results from these three maps provide consistency checks in relation to potential frequency-dependent foreground contamination. A total of 343,647 galaxies are used as tracers to identify and locate galaxy groups and clusters from which the kSZ signal is extracted using aperture photometry. We consider the impact of various aperture photometry assumptions and covariance estimation methods on the signal extraction. Theoretical predictions of the pairwise velocities are used to obtain best-fit, mass-averaged, optical depth estimates for each of five luminosity-selected tracer samples. A comparison of the kSZ-derived optical depth measurements obtained here to those derived from the thermal SZ effect for the same sample is presented in a companion paper.

Figure 1

Two null tests conducted on the pipeline for each galaxy tracer luminosity sample: Top: mean and standard deviation of the pairwise estimator derived from the average sample covariance for the DR4 ILC 560 noise simulations. Bottom: null test applied on the DR5 f150 map. Aperture photometries are taken from our science dataset for all luminosity bins; however, the temperature decrement values are randomly shuffled which removes the pairwise signal. Error bars show one sigma uncertainty inferred with bootstrapping temperature decrements. The ${\chi }^2$ and the probabilities to exceed it are also given for each test.

Figure 2

Pairwise velocity correlations for the DR4 ILC (black circle), DR5 f150 (blue cross) and DR5 f090 (orange square) maps for sources in the five luminosity-selected galaxy tracer samples, from top to bottom: $L43D$, $L61D$, $L43$, $L61$ and $L79$. Error bars show $1\sigma$ bootstrap uncertainties.

Figure 3

The normalized likelihood for $\overline{\tau }$ estimates for the DR5 f150 (blue, full), DR5 f090 (orange, dashed) and DR4 ILC (black, dotted) maps for each of the five luminosity-derived tracer samples using the boostrap uncertainties.

Figure 4

The extracted pairwise signal for the DR4 ILC (black, lower) and DR5 f090 (orange, middle) and DR5 f150 (blue, upper) maps for the three cumulative luminosity-selected galaxy tracer samples, $L43$ (left), $L61$ (center), $L79$ (right), overlaid with the theoretical pairwise velocity model using the Planck best-fit cosmology corresponding to the best-fit $\overline{\tau }$ value and $1\sigma$ boostrap-derived uncertainties.

Figure 5

Comparison of the impact of a constant (red) versus redshift-varying (blue)] aperture size in the pairwise momentum estimation, $\widehat{p}$, for the DR4 ILC map and the DR15 $L43$ sample.

Figure 6

Comparison of the impact of the redshift smoothing factor in the aperture photometry estimation for the DR4 ILC map and the DR15 $L43$ sample. Top: uncertainties introduced in ${\widehat{p}}_{\mathrm{kSZ}}$ when varying ${\sigma }_z=0.01$, 0.02, 0.05 are compared with $1\sigma$ uncertainties from noise simulations (assuming ${\sigma }_z=0.01$). Bottom: individual ${\overline{T}}_{\mathrm{AP}}$ realizations for each case.

Figure 7

Comparison of the impact of step-by-step changes in the aperture photometry assumptions in pairwise momentum estimation, $\widehat{p}$ obtained from those used in DB17 to this work. The stepwise changes in analysis and data are given in Table 3 for samples with a $L>7.9\times {10}^{11}{L}_{\odot }$ luminosity cut. The gray shaded region shows the $1\sigma$ boostrap-derived uncertainties for the DR5 f150 analysis.

Figure 8

Comparison of the standard deviation of 560 DR4 ILC noise simulations (“sim”) (green) to the uncertainties obtained from the jackknife (“jk”) (blue) and bootstrap (“bs”) (orange) resampling of a single noise simulation for (top) L43 and (bottom) L79 galaxy tracer samples.

Figure 9

Comparison of the distribution of ${\chi }^2$ values obtained from 560 simulation realizations evaluated on covariances inferred using: sims (black), jackknife (“jk”) (blue) and bootstrap (“bs”) (orange) resampling of a single sim for the DR4 ILC (top) L43 and (bottom) L79 samples versus the theoretical prediction for ${\chi }^2$ for 19 degrees of freedom (equal to the number of spatial separation bins).

Figure 10

The pairwise correlation matrix, for the DR4 ILC map across the 19 spatial cluster separation bins for the L43 galaxy tracer sample, derived from (top) the covariance across 560 noise sims, (center) the bootstrap and (bottom) jackknife resampling.