A Semiblind PCA-based Foreground Subtraction Method for 21 cm Intensity Mapping

Abstract The principal component analysis (PCA) method and the singular value decomposition (SVD) are widely used for foreground subtraction in 21 cm intensity mapping experiments. We show their equivalence, point out that condition completely clean separation of foregrounds cosmic signal using PCA/SVD is unrealistic. propose a PCA-based method, dubbed “singular vector projection (SVP)” which exploits priori information left and/or right vectors foregrounds. demonstrate with simulation tests this new, semiblind can reduce error recovered by orders magnitude, even if only largest few modes exploited. SVP estimators provide effective approach observations to remove uncover physics signal.