CMB Lensing

The large scale structure in the universe deflects the cosmic microwave background photons. The typical deflection is a couple of arcminutes, but the deflection field is coherent over a large range angular scales, up to about a few degrees. The deflections smooth acoustic features in the primary cmb anisotropy, moves power from large scales in the CMB to the damping tail and generates non-Gaussianity.

With my advisor, Professor David Spergel, I study the effect and its implications to cosmology both analytically and through numerical simulations.

I also intend to look for this effect in the data from the Atacama Cosmology Telescope (ACT) project.

Statistical Gravitational Lensing

With Professor Jeremiah P. Ostriker, I have investigated how the probability of strong lensing of a distant source by intervening structure depends on the cosmology. We came up with a way to analytically generate the probability density function of the dark matter convergence from a given input cosmology. The results have been published in the astrophysical journal, 645, 1 (2006).


CMB Power Spectrum Estimation

With Amir Hajian and David Spergel, I have proposed a new method for estimating the power spectrum of high resolution CMB maps. At the resolutions of ACT and the South Pole Telescope (SPT) the CMB maps will be contaminated with a large number of point sources which need to be masked. Such masking and also hard edges lead to aliasing of power, so that the mode-coupled power spectrum is highly biased at high multipoles. This bias leads to large error bars on the final power spectrum when standard decorrelation techniques are used. We came up with a real space operation called pre-whitening which effective reduced aliasing of power. Also, we generalized a technique, usually used in signal processing and known as the Adaptive Multitaper Method (AMTM) to two dimensions to deal with aliasing due to hard edges. With these methods, we showed that at multipoles above 2000 we were able to reduce the error bars over the standard methods by a large factor.