Descent, Seminormalization, and Conductors of Sparse Profile Images

Abstract

We develop a descent theory for sparse factorization-profile images, asking which functions on a normalization descend to the actual image. Whole-fiber completions yield a Fourier-zero classification for homogeneous masks, with a sharp stable converse, and a Fourier-line classification for mixed aperiodic profiles in the stated marked stable ranges. For homogeneous profiles, bounded integer relations determine geometric fibers and seminormalization, while explicit semigroup models compute actual image algebras, conductors, and depth in substantial families. For singleton--complement profiles, exact one-jet conditions determine the finite-output algebra, conductor, and projective descent through the first nonnormal boundary; the local type is computed in the smooth-boundary range and at the total-zero point of the first nonnormal layer. We also prove a saturated root-of-unity specialization theorem and give separating examples showing that normalization, natural polarizations, point fibers, seminormalization, and even the source conductor need not determine the actual image.