Title: Cooling Improves Cosmic Microwave Background Map-Making When Low-Frequency Noise is Large

Abstract: In the context of Cosmic Microwave Background data analysis, we study the solution to the equation that transforms scanning data into a map. As originally suggested in "messenger" methods for solving linear systems, we split the noise covariance into uniform and non-uniform parts and adjust their relative weights during the iterative solution. With simulations, we study mock instrumental data with different noise properties, and find that this "cooling" or perturbative approach is particularly effective when there is significant low-frequency noise in the timestream. In such cases, a conjugate gradient algorithm applied to this modified system converges faster and to a higher fidelity solution than the standard conjugate gradient approach. We give an analytic estimate for the parameter that controls how gradually the linear system should change during the course of the solution.