Anton Baleato Lizancos
The cosmic microwave background (CMB), relic radiation emitted when the Universe was only 380,000 years old, is the oldest light we can observe. Hidden in the curl pattern of its polarisation is information about the physics of the Universe much earlier on, when it was but a trillionth of a second old. Naturally, the curl-type polarisation of the CMB has become the holy grail of modern Cosmology. Unfortunately, it is obscured by a spurious polarisation curl induced by gravitational lensing of CMB photons by the large-scale structure of the Universe. In this poster, we explain common methods to "clean" the lensing contamination, what is usually known as "delensing". These efforts entail estimating the projection of the matter distribution on the sky, either by reconstructing it from the statistical imprint left by lensing on the CMB fields themselves, or by mapping the galaxies that trace the matter. With these estimates in hand, the lensing-induced deflections can be partially undone. In this poster, we present recent work that identifies possible biases to the delensing procedure due to 1) correlations between CMB polarisation fields used for lensing reconstruction and the polarisation fields we ultimately wish to delens and 2) residual galactic dust left over in far-infrared maps used for delensing. We also explore limitations of the commonly-used, leading-order template delensing method, and explain the surprising result that it is advantageous to build it using lensed, rather than delensed, E-modes. Finally, we showcase ongoing efforts to model biases to the CMB lensing power spectrum and cross-correlations due to extragalactic emission from clusters and galaxies. We present very fast code that enables marginalisation over these biases.
This can be mitigated by forming a leading-order B-mode template by convolving E-mode observations with a proxy of the lensing potential, and subtracting them from observations (delensing).
We show that 1) corrections to the leading-order calculation of the lensing B-mode power spectrum only enter at the O(1)% level because of extensive cancelations between large terms at next-to-highest order, 2) that these cancellations disappear when the template is built from unlensed or delensed E-modes, so the residual lensing floor is of O(10)% of the original power and 3) that new cancellations arise when lensed E-modes are used in the template, bringing the lensing floor to O(1)% in practical applications of this method.
IMPACT OF INTERNAL-DELENSING BIASES ON PRIMORDIAL B-MODE SEARCHES
The quadratic estimator (QE) (Hu & Okamoto 02) reconstructs the lensing potential from off-diagonal correlations between pairs of CMB fields.
In the near future, the EB QE will dominate reconstructions. Any overlap in modes between the B-field to be delensed and the B-field from which the reconstruction is derived leads to a suppression of delensed power beyond that associated with a mitigation of lensing (Teng et al. 11). Crucially, the variance of the delensed field is also suppressed. Could this help better constrain primordial B-modes?
We model the bias and show that its presence necessarily degrades the signal-to-noise on a primordial component. The bias can be mitigated by removing the largest angular scales from the B-field to be fed to the QE. We show that it is in general advantageous to do this rather than modelling or renormalising the bias.
DELENSING B-MODES WITH THE CIB: THE IMPACT OF FOREGROUNDS
For near-future experiments, the best estimate of the CMB lensing potential entails co-adding internal reconstructions the cosmic infrared background (CIB) (Sherwin & Schmittfull 15).
Galactic dust and CIB have similar SEDs, so disentangling them is difficult. We worry that dust/CIB residuals in CMB maps can couple with residuals in CIB intensity maps used as matter tracers for delensing and bias the power spectrum of delensed B-modes (and hence ).
We show that bias from higher-point functions of CIB itself is negligible.
Bias due to <BEI> bispectrum of galactic dust is non-negligible. However, we show that a simple mitigation technique (deprojecting dust from E-modes in an ILC) is very effective while incurring little loss in delensing efficiency.
CMB LENSING BIASES FROM GALAXIES AND CLUSTERS
Measurements of the lensing convergence power spectrum and cross-correlations with other, lower-redshift matter tracers contain a wealth of cosmological information.
However, when the lensing reconstruction is derived from a TT QE these spectra are biased by the thermal Sunyaev-Zeldovich (tSZ) effect from clusters and the CIB (Van Engelen et al. 14, Osborne et al. 14).
We strive to analytically calculate the biases as a function of experimental sensitivity, resolution, point-source masking, etc, using a halo model prescription, in order to enable marginalisation over them in actual analyses.
We’ve produced a very fast code that does the lensing reconstruction step in 1D using FFTlog, and another which does it in 2D and is thus slower, but it can incorporate cluster tri-axiality and realistic noise filtering.