Precision measurements of cosmic microwave background ( CMB ) polarization require extreme control of instrumental systematics .
In a companion paper we have presented cosmological constraints from observations with the BICEP2 and Keck Array experiments up to and including the 2015 observing season ( BK15 ) , resulting in the deepest CMB polarization maps to date and a statistical sensitivity to the tensor-to-scalar ratio of \sigma ( r ) = 0.020 .
In this work we characterize the beams and constrain potential systematic contamination from main beam shape mismatch at the three BK15 frequencies ( 95 , 150 , and 220 GHz ) .
Far-field maps of 7,360 distinct beam patterns taken from 2010–2015 are used to measure differential beam parameters and predict the contribution of temperature-to-polarization leakage to the BK15 B -mode maps .
In the multifrequency , multicomponent likelihood analysis that uses BK15 , Planck , and WMAP maps to separate sky components , we find that adding this predicted leakage to simulations induces a bias of \Delta r = 0.0027 \pm 0.0019 .
Future results using higher-quality beam maps and improved techniques to detect such leakage in CMB data will substantially reduce this uncertainty , enabling the levels of systematics control needed for BICEP Array and other experiments that plan to definitively probe large-field inflation .