ACT (AdvACT) DR6 Lensing Maps

Version: v1
Date: Nov 1, 2023
- Mathew Madhavacheril (
- Frank Qu (
- Gerrit Farren (

This directory contains gravitational lensing convergence maps (in Equatorial coordinates) from measurements of the CMB made by the Atacama Cosmology Telescope Data Release 6.

NOTE: This tarball on LAMBDA only includes the data maps and not the accompanying simulations. The full release including the suite of 400 simulated reconstructions can be accessed on the NERSC filesystem at the location /global/cfs/projectdirs/act/www/dr6_lensing_v1/ or through this http url.

Tutorials: We have prepared Colab Python notebook tutorials at this http url covering various aspects of using these maps.

ACT DR6 CMB lensing map release


If you use the data products in this directory, please cite:
1. M. Madhavacheril et al, arxiv:2304.05203, ApJ
2. F. Qu et al, arxiv:2304.05202, ApJ

If you use the accompanying simulations, please also cite the
following paper in addition to the above two:
3. G. Farren et al, arxiv:2309.05659


We provide several map variants in the directories `maps/{$VARIANT}`. We recommend using `baseline` for your cosmology or astrophysics inference. These variants are described briefly below, but we recommend referring to Madhavacheril et al and Qu et al for details.

  1. baseline:
    The baseline map used in the ACT DR6 lensing cosmology analysis, made
    from all temperature and polarization data, and using a profile-hardening
    estimator for foreground mitigation.
  2. ponly:
    A map that only uses polarization data
  3. tonly:
    A map that only uses temperature data
  4. cibdeproj:
    A map that uses CIB deprojection in lieu of profile-hardening as an
    alternative method for foreground mitigation
  5. f090:
    A map that only uses 90 GHz data from ACT
  6. f150:
    A map that only uses 150 GHz data from ACT
  7. f090_tonly:
    A map that only uses 90 GHz temperature data from ACT
  8. f150_tonly:
    A map that only uses 150 GHz temperature data from ACT
  9. galcut040:
    A map that is reconstructed on a more restricted area that discards
    more regions closer to the Galactic plane
  10. galcut040_ponly:
    A map that is reconstructed on a more restricted area that discards
    more regions closer to the Galactic plane, using only polarization
  11. diff_f150_f090:
    A map reconstructed from a difference of maps at 150 GHz and 90 GHz.
    This is a null map; it is not expected to contain any cosmological signal.
  12. diff_f150_f090_tonly:
    A map reconstructed from a difference of temperature maps at 150 GHz and 90 GHz.
    This is a null map; it is not expected to contain any cosmological signal.
  13. curl:
    A map that reconstructs a curl component, which is expected to be
    a null map with no cosmological signal.


Products associated with each variant are provided in sub-directories. For each variant {$VARIANT} from the list above, we provide the following.

  1. Map alm modes: {$VARIANT}/kappa_alm_data_act_dr6_lensing_v1_{$VARIANT}.fits
  2. Mask: {$VARIANT}/mask_act_dr6_lensing_v1_healpix_nside_4096_{$VARIANT}.fits
  3. Filter: {$VARIANT}/kappa_filter_act_dr6_lensing_v1_{$VARIANT}.txt
  4. Noise per mode: {$VARIANT}/N_L_act_dr6_lensing_v1_{$VARIANT}.txt
  5. Simulated map alm modes: {$VARIANT}/simulations/kappa_alm_sim_act_dr6_lensing_v1_{$VARIANT}_{$i}.fits

Products: Map

The map alm modes are the spherical harmonic modes of the CMB lensing convergence (kappa) provided in the healpy indexing scheme for alm's. They correspond to modes of a map described in Equatorial coordinates, as opposed to the Galactic coordinates used by Planck products. These spherical harmonic modes may be be projected to a pixelization of your choice, although note that we provide a mask (described below) in HEALpix format.

Products: Filter

The modes are only non-zero for 2 ≤ L ≤ 3000, as explicitly described in the filter file (which lists ones for multipoles in this range, and zeros elsewhere). However, not all modes within that range may be reliable for use in analysis. We refer the user to the references above and encourage the use of simulations to assess the validity of the multipole range used, which will depend on the exact application.

Products: Noise per mode

The noise per mode file provides an N_L curve appropriate for use in forecasting for cross-correlations. Due to the use of a split-cross-correlation-based estimator, this noise curve is *not* an estimate of the N_0 bias seen when taking the power spectrum of the provided maps. The noise is an estimate over the full footprint of the map; estimates over smaller footprints may differ substantially, and so this curve should only be used for approximate forecasting.

Products: Simulations

In order to facilitate interpretation of the lensing maps and their noise properties, we provide a set of 400 simulations for each variant. These simulations are made as follows: we generate input lensing convergence fields as Gaussian random fields from a fiducial lensing convergence spectrum. These are used to lens Gaussian CMB temperature and polarization maps drawn from a fiducial unlensed CMB temperature and polarization power spectra. These CMB maps are then processed through the ACT simulation pipeline to create mock ACT data-sets, which are then processed through the CMB lensing reconstruction pipeline to produce output lensing convergence reconstructions. We provide the input lensing convergence fields in `sim_inputs/kappa_alm/` and the simulated reconstructions corresponding to these for each variant in `maps/{$VARIANT}/simulations/`. The theory spectra used to generate the inputs to the simulations along with their CAMB configuration file are provided (in standard CAMB format) in `sim_inputs/theory`. These are identical to those used in the earlier ACT DR4 release.

These reconstructed lensing convergence maps have the same interpretation as the data map for that variant. In combination with the inputs in the `sim_inputs/kappa_alm/` the reconstructed simulations for each variant should be used to (a) estimate a transfer function for cross-correlations (see Mask section below) and (b) estimate covariance matrices if needed. We reiterate that the inputs to the simulations do not come from N-body or large-scale structure simulations but are simply Gaussian realizations generated from fiducial theory spectra, and so they are limited to the above applications.

Products: Mask

We provide a mask in HEALpix format with nside=4096 in Equatorial coordinates. This mask was applied to the CMB map before use in the lensing quadratic estimator (which involves two powers of the CMB map). It is apodized and contains ones in most of the footprint, which transition to zero smoothly near the edges. For a real-space analysis, we recommend rejecting points where this mask is different from 1 (e.g. mask<0.99) and forward modeling the effect of the filter on your signal from harmonic space to real-space. The default filter (as specified in the filter file) is 2 ≤ L ≤ 3000, however, you will likely need to exclude more modes for robust results.

For a harmonic-space analysis, the effect of the mask must be deconvolved (e.g. using the MASTER algorithm). This is not straightforward due to the mask having been applied before applying a quadratic function to the CMB map. We therefore highly recommend the following procedure: (a) in a cross-correlation involving the ACT map K (with the ACT mask MK) and an external large-scale structure map L (with mask ML, where the observed map L is linear in ML), use the MASTER algorithm with mask MK2 associated with K and ML associated with L. (b) repeat this procedure on the provided 400 simulations by doing a similar cross-correlation of the reconstruction simulations in `maps\{$VARIANT}\simulations\` (masked with MK and corrected with MK2 in the MASTER code) with the input lensing convergence fields in `sim_inputs/kappa_alm` (masked with ML and corrected with ML in the MASTER code).

The procedure in (b) is essential to estimate a small residual transfer function (at the few percent level) that will exist because deconvolution with MK2 is an approximation and because the mask ML may select a region very different from the region MK that was used to correctly normalize the map.

Back To ACT

A service of the HEASARC and of the Astrophysics Science Division at NASA/GSFC
Goddard Space Flight Center, National Aeronautics and Space Administration
HEASARC Director: Dr. Andrew F. Ptak
LAMBDA Director: Dr. Thomas M. Essinger-Hileman
NASA Official: Dr. Thomas M. Essinger-Hileman
Web Curator: Mr. Michael R. Greason