Simple characteristic decomposition of polynomial sets (Q6950602)

The paper under review uses triangular sets and Gröbner bases (two important tools in computer algebra) as a foundation for its new method. The paper introduces algorithms for decomposing polynomial systems into characteristic pairs with simple W-characteristic sets over fields of both characteristic zero and finite characteristic. A characteristic pair \((G, C)\) consists of a reduced lexicographical Gröbner basis \(G\) of an ideal and the minimal triangular subset \(C\) of \(G\), called its W-characteristic set. This pair captures two complementary algebraic representations of the same ideal -- one through Gröbner bases and one through triangular sets. Given a finite set of polynomials \(F\), a finite set \(\{(G_1, C_1), \cdots, (G_t, C_t)\}\) of characteristic pairs is called a characteristic decomposition of \(F\) if the following equality holds:\N\[\NZ(F) = \bigcup_{i=1}^t Z(G_i) = \bigcup_{i=1}^t Z(C_i / \text{ini}(C_i)) = \bigcup_{i=1}^t Z(\text{sat}(C_i)).\N\]\NA simple W-characteristic set (or simply a simple set) is a triangular set whose successive polynomials are square-free and whose saturated ideal is radical, ensuring that the decomposition corresponds to distinct algebraic components without multiplicity. When the equality \(\operatorname{sat}(C)=\langle G\rangle\) holds, the pair \((G,C)\) is called a strong characteristic pair. A characteristic decomposition is said to be strong or simple if every characteristic pair in the decomposition is strong or simple, respectively. In the first part of the paper, the authors present an algorithm over fields of characteristic zero that uses Lazard's lemma to test the radicality of ideals without explicitly computing radicals, which significantly improves efficiency. In the second part, an alternative algorithm is proposed for finite fields, where the lemma cannot be applied; in this case, radicals are computed directly with the help of existing algebraic software. Both algorithms are carefully proved to terminate correctly and to produce decompositions satisfying the expected zero relations. Several worked examples illustrate the intermediate steps, and an extensive experimental section provides performance results on standard benchmark systems, demonstrating that the new methods are both effective and competitive with existing approaches.