Two letters from Bernard Malgrange on the nonlinear differential Galois theory (Q6607750)

It is a well-known fact that a number of properties taken for granted in linear differential systems disappear or become diffuse in more general situations where linearity is not assumed. One of these properties is the existence of a differential Galois group (and, generally speaking, a consistent Picard-Vessiot theory) attached to said systems.\N\NThe present paper comprises a relatively recent short epistle exchange between the late Bernard Malgrange and two other world-class experts on the subject (Jean-Pierre Ramis and Guy Casale) about the two main attempts to generalize differential Galois theory to the non-linear realm. These two attempts, due to Malgrange and the late Hiroshi Umemura, were concomitant in time and shared a substantial common ground.\N\NIn his first letter to Ramis, Malgrange starts by clarifying the aforementioned common ground: differential equations as defined by fields \(\xi\) on smooth complex algebraic varieties \(X\). He then proceeds to introduce the Galois pseudogroup of \(\xi\) and its envelope--the former of which was not considered by Umemura. In order to do this, given a Zariski open set \(U\) (i.e. the complement of an algebraic subset) of \(X\), he first introduces the order-\(k\) jets \(J_0^k\left(U\right)\) as well as the order-\(k\) principal bundle \(R_k\left(U\right)\) over \(U\) having \(\Gamma_k\) as their group of \(k^{\mathrm{th}}\) order invertible jets of \(\left(\mathbb{C}^n,0\right)\). He considers a reduced subgroupoid of \(J_0^k\left(U\right)\) and the germs thereof via the usual chart-induced equivalence relation (albeit restricted to a smaller open set \(V\subset U\)) in a vicinity of a generic point \(x\in X\). A pseudogroup on \(U\) is subsequently defined as a collection of such subgroupoids defined by a radical ideal. A germ of pseudogroups naturally possesses a Lie algebra. The envelope of \(\xi\) is then defined to be the minimal germ of pseudogroups rendering \(\xi\) a solution to its own Lie algebra, and its (germ of) Galois pseudogroup is the minimal germ of pseudogroups such that all multiples of \(\xi\) are solutions to its Lie algebra.\N\NThat the latter definition (i.e. that of Galois pseudogroups) only depends on the envelope of \(\xi\) rather than \(\xi\) itself plays a preponderant role in the exposition, given Umemura's eschewance of the Galois pseudogroup and exclusive consideration of the envelope (denoted \(P_k\) at its \(k^{\mathrm{th}}\) order by Malgrange).\N\NGiven the rational function field \(K=\mathbb{C}\left(X\right)\) and \(H=K^\xi\) the field of first integrals of \(\xi\), a germ of equivalence defining \(P_0\) is given by the prime ideal \(I=\mathcal{I}\left(H\right)\subset L=K \otimes_H K\) generated by all \(a\otimes 1-1\otimes a\), \(a\in H\). The primality of \(I\) implies \(L\) is an integral domain and is a consequence of the fact that \(K^\xi\) is algebraically closed in \(K\). For higher \(k\ge 1\), the situation is analogous by replacing \(\xi\) with its \(k\)-jet prolongations stable under \(\Gamma_k\). The author then proceeds to introduce Umemura's formulation, based on the Zariski closure of the holonomy.\N\NThe next section concerns the global or regular theory. The notion ``regular'' in this context means being locally expressible as a rational function, i.e. a ratio of polynomials with non-vanishing denominators. The space \(\mathcal{O}_k\left(U\right)\) of functions regular on \(J_0^k\left(U\right)\) satisfies \(\mathcal{O}_k\left(U\right) \subset \mathcal{O}_k\left(V\right)\) if \(V\subset U\), and germs \(P_k\) are defined as ideals of rings \(\mathcal{O}_k\). This leads to the Zariski closure \(\overline{P}_k\) of \(P_k\); \(\overline{P}=\left(\overline{P}_k\right)\) fails to be a pseudogroup in general. In view of this ostensible incompleteness, the letter then refers to different contexts in which \(\overline{P}\) appears naturally, posits the immanent rational character of differential Galois theory, and disquires on the apparent lack of exploration of this potential in Umemura's work. The letter then ends with some remarks elaborating further on the potential implications of this theory.\N\NThe second chapter contains a letter by the author to Guy Casale, wherein Malgrange presents an alternative proof for a 2009 result by Casale about Morales-Ramis theory. Taking a good representative of the Lie pseudogroups (i.e. a representative on a Zariski-dense open set of a smooth, singularity-free variety \(X\)) as his choice of definition for Galois pseudogroups, Malgrange establishes a relationship between the ideal \(\mathcal{I}\) defining the Umemura ring as a quotient of \(A\otimes_{\mathbb{C}}A\) (\(A\) being the ring of regular functions over \(X\)), and the degree-zero part of the Galois pseudogroup of vector field \(\xi\). The analog case in which \(A\) is replaced by its field of fractions \(K\) is then studied, and the author characterizes the corresponding ideal \(\mathcal{I}\) both as the localization of an appropriate finitely-generated ideal \(\mathcal{I}'\) of \(A\otimes_{\mathbb{C}} A\), and as defined by a relation of equivalence \(R\) on \(X\setminus Z\), \(Z\) being the set of singularities of all generators of \(\mathcal{I}'\). It is the latter characterization (in terms of \(R\)) and a simple application of Hilbert's Nullstellensatz, that allow the author to provide Casale's theorem with a new proposed proof. A final remark throws interesting light on the possible obtention of the higher variational complex of any system in terms of a restriction of the above constructions to an algebraic orbit of \(\xi\), thereby providing a potential link between two major branches of study of the extended differential Galois theory of dynamical systems.