Bias-Hardened CMB Lensing

We present new methods for lensing reconstruction from CMB temperature fluctuations which have smaller mean-field and reconstruction noise bias corrections than current lensing estimators, with minimal loss of signal-to-noise. These biases are usually corrected using Monte Carlo simulations, and to the extent that these simulations do not perfectly mimic the underlying sky there are uncertainties in the bias corrections. The bias-hardened estimators which we present can have reduced sensitivity to such uncertainties, and provide a desirable cross-check on standard results. To test our approach, we also show the results of lensing reconstruction from simulated temperature maps given on 100 deg^2, and confirm that our approach works well to reduce biases for a typical masked map in which 70 square masks each having 10 arcminute on a side exist, covering 2% of the simulated map, which is similar to the masks used in the current SPT lensing analysis.