An Efficient Algebraic Algorithm for the Geometric Completion to Involution

 We describe an adaption of a differential algebraic completion algorithm for linear systems of partial differential equations that allows us to deduce intrinsic differential geometric information like the number of prolongations and projections needed for the completion. This new hybrid algorithm represents a much more efficient realisation of the classical Cartan–Kuranishi completion than previous purely geometric ones. A classical problem in geometric completion theory is the existence of δ-singular coordinate systems in which the algorithms do not terminate. We develop a new and a very simple criterion for δ-singularity based on a comparison of the Janet and the Pommaret division. This criterion can also be used for the direct construction of δ-regular coordinates. Received: July 28, 2000; revised version: October 16, 2001