Macaulay inverse systems and Cartan-Kahler theorem

Abstract

During the last months or so we had the opportunity to read two papers trying to relate the study of Macaulay (1916) inverse systems with the so-called Riquier (1910)-Janet (1920) initial conditions for the integration of linear analytic systems of partial differential equations. One paper has been written by F. Piras (1998) and the other by U. Oberst (2013), both papers being written in a rather algebraic style though using quite different techniques. It is however evident that the respective authors, though knowing the computational works of C. done during the first half of the last century in a way not intrinsic at all, are not familiar with the formal theory of systems of ordinary or partial differential equations developped by D.C. Spencer (1912-2001) and coworkers around 1965 in an intrinsic way, in particular with its application to the study of differential modules in the framework of algebraic analysis. As a byproduct, the first purpose of this paper is to establish a close link between the work done by F. S. Macaulay (1862-1937) on inverse systems in 1916 and the well-known Cartan-K{\"a}hler theorem (1934). The second purpose is also to extend the work of Macaulay to the study of arbitrary linear systems with variable coefficients. The reader will notice how powerful and elegant is the use of the Spencer operator acting on sections in this general framework. However, we point out the fact that the literature on differential modules mostly only refers to a complex analytic structure on manifolds while the Spencer sequences have been created in order to study any kind of structure on manifolds defined by a Lie pseudogroup of transformations, not just only complex analytic ones. Many tricky explicit examples illustrate the paper, including the ones provided by the two authors quoted but in a quite different framework.