Hypergeometric solutions of linear difference systems (Q6890940)

The paper develops algorithms for finding hypergeometric solutions of linear difference systems, generalising Petkovšek's method from single equations to full systems. A solution is viewed as a hypergeometric term multiplied by a primitive polynomial vector. The authors begin with Bronstein's idea of bounding candidates by examining the denominators of the system matrix and its inverse, but they then refine this approach substantially. They introduce the use of unramified generalised exponents, computed through super-reduction, to restrict the possible hypergeometric factors. Compatibility conditions derived from these exponents link the slope, constant factor, and index of potential solutions. The authors also define the notions of type and local type to group equivalent candidates and select minimal representatives, preventing duplicate solutions. The final algorithm efficiently produces polynomial solutions and scales well to high-dimensional systems, supporting applications such as the factorisation of difference operators.