Chicago-2016: Lensing

From CMB-S4 wiki
Jump to: navigation, search

Summary of Science Book Chapter 7 - CMB Lensing

Session Leader: Neelima Sehgal

File:CMBS4 CMBLensing.pdf

What we have learned so far

1.) Staying below ~4uk-arcmin (regular lens) and ~1uK-arcmin (halo lens) avoids temperature atmospheric noise and many foregrounds.

2.) Foreground-cleaned map for delensing needs to be deeper than foreground-cleaned map for primordial B-modes.

3.) Bias on lensing translates directly into a bias on r. Single dish that can do low and high ell is most efficient and most conservative from lensing view.

4.) ~1.5’ resolution should be fine for all lensing areas, which is a 5-meter dish at 150 GHz. Caveat is if we need to foreground clean with synchrotron channel at high resolution.

Next Steps

1.) Re-do non-r forecasts with realistic noise curves for temperature, and think of clever ways to deal with atmosphere if considering wide survey.

2.) Figure out impact of polarized Galactic foregrounds on lensing and how many frequencies are needed to clean lensing map from pol.

3.) Figure out if adding low-redshift, large-scale structure data (such as from optical weak lensing or clusters) to cmb lensing and the primary cmb can mitigate the tau issue for neutrino mass.

Notes from session

Science Book Chapter 7: CMB Lensing

Neelima Sehgal

Lensing/Delensing intro.

  • We measure lensed T, E, B; we want unlensed T, E, B, and \kappa
  • We constrain Neff, mnu, w, r

Areas of lensing.

  • Large-scale lensing
  • Small-scale lensing (cluster lensing)
  • Cross correlations
  • Calibrate LSST shear

Noise requirements.

  • Take away 1: Go to low noise to rely on pol rather than temperature.
    • Large-scale lensing: crossover at 4 uK-arcmin
    • Small-scale lensing: crossover at ~1 uK-arcmin
  • Next step 1: re-do non-r forecasts with realistic TT noise.
  • Cross-correlations: less sensitive to atmosphere noise maximize overlap
  • Adrian: multicolor single-pixels suppress atmosphere, might be promising.
  • Bill: Depends on scan speed, with fast scan-speed it may not be a problem.

Iterated delensing -- sensitivity vs area.

    • Take-away 2: Foreground-cleaned map should be deeper than the foreground-cleaned map for primordial B modes
    • Next step 2: How many frequencies do we need to clean out this lensng map? This is because small-scale Galactic foregrounds can bias lensing reconstruction.
    • 1% bias in lensing B modes gives 10% bias in r.
    • Take-away 3: bias on lensing translates directly to a bias on r.
    • If we need 3 widely spaced freqs, diving into 2 or 3 via multi-chroic gives 7 frequencies.


  • Significant gain from high-resolution for small deep patches.
  • Modest gain for LSS/mnu science.
  • Significant gains for mass resolution for cluster lensing.
  • Take-away 4: 1.5' resolution should be fine for all lensing areas, a 5-meter dish at 150 GHz. Caveat: we might need synchrotron at high resolution.


  • John Ruhl: Do we assume unique lmax for the forecasting? Neelima: 3000 for T, 5000 for P
  • Jaques: Why deeper on small-scale than large-scale survey? Neelima: Blake's plot shows this, if you take ClBB_residual to be an effective white noise level in uK-arcmin.
  • Steve Allen: You also need the high angular resolution to find the clusters, not just lensing. Neelima': right -- what is tradeoff in selection function for 1.5' vs 1'? Brad: you lose clusters more quickly at high redshift with beam scale.
  • Suzanne: The numbers for 5m dish for 1.5' for 150 GHz are rough.
  • Tom: Biases on r, it seems bad to have 10% bias if there is a 1% lensing bias. It has an ell dependent shape, can't we fit it? Blake: yes, but there can be fluctuations around the shape that can look like r. Neelima: yes, but it can inflate our error bar on r, in principle
  • John Kovac: Whatever level of removal you try, there is potential for bias on both delensing and gaalactic foregrounds.