Work Update: EDF-S



Cail Daley

CosmoStat Group Meeting

September 23, 2025

Outline

  • Background on Euclid Q1 and EDF-S

  • EDF-S CMB lensing motivation

  • Current status of lensing analysis

  • Next steps & timeline

Why this presentation may not be the best

Euclid Q1 & EDF-S

  • Euclid’s “Quick Data Release 1” (Q1) is the first public data from Euclid, and includes 3 fields (EDF-S, EDF-F, EDF-N) totaling 63 deg².

    • More observations will be conducted to turn the Q1 fields in Deep Fields (50x deeper than the Wide Survey) for legacy science.

  • SPT observed EDF-S last fall, and did a data release of catalogs, temperature maps, and associated data products at the beginning of June.

CMB Lensing with EDF-S

  • EDF-S will become a legacy deep field, and CMB lensing lensing maps will be a useful data product for the community.

  • While Q1 data was nominally not intended for cosmology, the galaxy clustering signal is very strong. Giulio Fabbian has been leading a 2x2pt galaxy clustering \(\ev{\delta_g\delta_g}\) & CMB lensing \(\ev{\delta_g \kappa_{\rm CMB}}\) analysis with ACT and Planck lensing maps.

  • The ACT cross-correlation SNR is about 6—we can do much better with SPT!

Mapmaking

Aman Chokshi

SPT takes data by performing constant-declination
scans across the sky:

  • 72 scans in a given 2-hour observation,
  • 12.5 arcmin elevation step between scans,
  • 0.3-0.6 arcmin\(^{-1}\) sampling rate,
  • ~14 000 useable detectors


Fundamental data product: timestreams.

  • To get to a map: filter timestreams, bin into pixels.

Mapmaking

Source Templates

  • Using a higher mapmaking threshold of 20 mJy in combination with a source template to reduce number of holes in the map and filtering artifacts.

    • Source template: mock-observed map of beam-convoled sources made from the EDF-S source catalog.

  • Sources > 20 mJy and clusters will still be inpainted.

  • Only sources > 6 mJy are included in the template, to match the foreground power in the 2-year Gaussian lensing sims.

  • This approach has required several iterations, but seems to be working well.

Source Templates

Source Templates

Component Separation

Leverage frequency-dependence of noise and foregrounds to make
a cleaner estimate of the CMB: \[X_{\vbell}^{\rm MV} = \sum_{i} \mathbb W_{\vbell}^{i} X_{\vbell}^{i}.\]

Component Separation

  • I shared these data products with Victor over the summer and he kindly sent me back ML-seperated maps.

  • but I haven’t had a chance to look at them yet 😔

Lensing Reconstruction

  • Adapting pipeline from the 2-year winter flat-sky QE analysis.

  • So far only running on data; so no mean-field, Monte-Carlo response, etc.

Cross-correlations with Euclid

  • Galaxy density maps and ACT bandpowers kindly provided by Giulio.
  • Error bars calculated with NaMaster.

Summary & Next Steps

  • Data processing aspects largely wrapped up.
  • Lensing maps maps look sensible but amplitude seems to be 40-50% low,
    likely due to missing Monte Carlo correction.
  • Cross-correlation SNR ~20% higher than ACT
    (6.8 vs 8.2 for single tomographic bin).
  • Next steps:
    • Run on sims to get response, mean-field, and noise bias corrections.
      \(\implies\) some level of lensing power spectrum validation.
    • Map calibration (Pcal, T\(\to\)P, EB nulling…)
    • Try Julien Carron’s lensing reconstruction code

Timeline & Euclid DR1

  • Planning to have a draft this fall: currently halway through map calibration.
  • One paper on lensing maps and x-corr? Or two separate ones?
    • If the latter, could add a public shear catalog and do a mini 3x2pt
      (\(\delta_g\kappa_{\rm CMB}\), \(\delta_g\gamma\), \(\gamma \kappa_{\rm CMB}\))
  • Euclid CMBX RR2 Hands-On Meeting Oct 7-8 in Paris
  • DR1 EC/MoU proposal for a Euclid-SPT 3 \(\times\) 2pt accepted by respective committees (waiting for final ECEB approval).

Credit: Margherita Lembo

Thanks!