Symbolic summation in multivariate difference fields (Q6907732)

Let \((\mathbb{F},\phi)\) be an ordinary difference field with basic automorphism \(\phi\). A multivariate difference field extension \((\mathbb{F}(\alpha_{1},\dots,\alpha_{n}),\phi)\) of \((\mathbb{F}, \phi)\) is defined as a field extension satisfying the following conditions: (1) \(\alpha_{1},\dots, \alpha_{n}\) are algebraically independent over \(\mathbb{F}\); (2) \(\phi\) is an automorphism of \(\mathbb{F}(\alpha_{1},\dots, \alpha_{n})\), \(\phi|_{\mathbb{F}} =\) id, and \(\phi(\alpha_{1},\dots, \alpha_{n}) = (\alpha_{1},\dots, \alpha_{n})A\) for some invertible matrix \(A\in \mathbb{F}^{n\times n}\).\N\NThe paper under review develops an algebraic framework for symbolic summation in multivariate difference fields. Inspired by Karr's theory of \(\Pi\Sigma\)-fields for univariate indefinite summation, the authors extend these ideas to sequences satisfying linear recurrences of higher order and to summation problems involving several shift variables. The central goal is to decide whether a rational function \(f \in F(\alpha_1,\dots,\alpha_n)\) is \(\sigma\)-summable in a multivariate difference field \((F(\alpha_1,\dots,\alpha_n),\sigma)\), that is, whether there exists \(g\) such that\N\[\N\sigma(g) - g = f.\N\]\NThe first main contribution is a structural reduction. Under suitable assumptions (notably diagonalizability of the defining matrix of the automorphism), the authors show that a general multivariate difference field is isomorphic to one in which the automorphism acts diagonally,\N\[\N\sigma(\alpha_1,\dots,\alpha_n)=(\alpha_1,\dots,\alpha_n)\operatorname{diag}(\lambda_1,\dots,\lambda_n).\N\]\NThis reduction (Theorem 2.1 and its consequences) allows one to translate summation problems into a more tractable normal form. A detailed analysis of the constant field is provided (Theorem 2.5), expressed in terms of the exponent lattice generated by the multiplicative relations among the \(\lambda_i\).\N\NA second key ingredient is the introduction of the \textit{spread} \(\mathrm{Spr}_\sigma(p,q)\) of two polynomials, which measures how their \(\sigma\)-orbits intersect. The authors prove that the spread is a subgroup of \(\mathbb{Z}\) and provide an effective method for computing it via linear systems derived from exponent lattices (Proposition 2.10 and Algorithm 1). This yields a practical criterion for deciding equivalence under shifts and forms the computational backbone of the method.\N\NThe final part of the paper establishes an existence criterion for rational solutions of the difference equation. Using an orbital partial fraction decomposition with respect to \(\sigma\), the summability problem is reduced to independent problems in certain vector spaces associated with irreducible denominators. The main equivalence theorem (Proposition 3.5) shows that \(f\) is \(\sigma\)-summable if and only if certain reduced components are summable in lower-dimensional difference fields, ultimately reducing the problem to explicit multiplicative conditions.\N\NOverall, the paper provides a coherent algebraic theory and a concrete algorithmic procedure for symbolic summation in multivariate difference fields. It significantly extends the scope of Karr-type summation algorithms and offers both structural insight and computational tools for handling rational summation problems in several variables.