The Effects of Bandpass Variations on Foreground Removal Forecasts for Future CMB Experiments

TL;DR

This paper investigates how atmospheric and instrumental bandpass variations affect foreground removal in future CMB experiments, quantifies their impact on tensor-to-scalar ratio measurements, and establishes calibration accuracy requirements.

Contribution

It models atmospheric effects on bandpass variations and assesses their impact on foreground removal and r measurement accuracy in CMB experiments.

Findings

Uncertainties in band gain must be <2% for accurate r measurement.

Uncertainties in central frequency must be <1% to control bias.

Proper calibration can mitigate the impact of bandpass variations.

Abstract

Time-dependent and systematic variations in band gain and central frequencies of instruments used to study the Cosmic Microwave Background are important factors in the data-to-map analysis pipeline. If not properly characterized, they could limit the ability of next-generation experiments to remove astrophysical foreground contamination. Uncertainties include the instrument detector band, which could systematically change across the focal plane, as well as the calibration of the instrument used to measure the bands. A potentially major effect is time-dependent gain and band uncertainties caused by atmospheric fluctuations. More specifically, changes in atmospheric conditions lead to frequency-dependent changes in the atmospheric transmission which, in turn, leads to variations in the effective gain and central frequency of the instrument's bandpass. Using atmospheric modeling software and ACTPol bandpasses, we simulate the expected variations in band gain and central frequency for 20, 40, 90, 150, and 240 GHz bands as a function of precipitable water vapor, observing angle, and ground temperature. Combining these effects enables us to set maximum and minimum limits on the expected uncertainties in band gain and central frequency over the course of a full observing season. We then introduce the uncertainties to parametric maximum-likelihood component separation methods on simulated CMB maps to forecast foreground removal performance and likelihoods on the tensor-to-scalar ratio r. We conclude that to measure a $\sigma(r=0)$ ~ $10^{-3}$ with a bias on the recovered $r$ under control, the limit on the uncertainty in the relative gain and central frequency of the bandpass must be <2% and <1%, respectively. We also comment on the possibility of self-calibrating bandpass uncertainties.