Lensing Systematics - Plenary Summary

From CMB-S4 wiki
Jump to: navigation, search

Instrumental and modeling systematics for lensing: challenges, solutions and open questions

Instrumental systematics

  • Instrumental systematics (particularly beam systematics) can potentially mimic lensing
  • Some calculations (Su, Yadav, Zaldarriaga; Miller, Keating; CMBPol...) have been performed to quantify the effects of these systematics on lensing, assuming simple models

NewTable.png

  • For an S4-type experiment, it seems that the requirements on lensing are much less stringent than for going after r (Meng Su: even control at the level of r=0.01 is 10 x stronger than what is needed for lensing)
  • In addition, we can actually form dedicated estimators for measuring the levels of known systematics
  • CMBPol white paper gives similar result, that beam systematics can be characterized sufficiently well for high precision lensing.
  • We hope to make these estimates more realistic

Modeling systematics

1. Non-Gaussianity of lensing potential

  • so far, it was assumed that the lensing field was Gaussian. However, non-linear evolution imparts a non-zero bispectrum which affects the lensing power spectrum estimate (a four-point function):

Bispec.png

  • leads to biases of order 1 sigma for every L-bin of width 100 for temperature, 10x less for polarization

BispecBias3.png

  • must take these and other modeling systematics into account, probably with simulations

2. Lensing power spectrum modeling

  • lensing power itself must be accurate to 0.1% or better for all parameters - to what extent will we have emulators to do this?

Open Questions and Tasks

  • How can we make our systematics estimates more realistic, and what is missing in our estimates? How do we include this in forecasts?
  • Simulate full non-Gaussian lensing map, propagate through lens pipeline to estimate higher order biases.
  • Ensure our theory for the lensing power spectrum is sufficiently robust
  • How do issues propagate through delensing procedure? Lots of simulation and theory work needed.