The paper deals with a class of mathematical objects called A-hypergeometric systems, which are sets of differential equations whose solutions generalize classical special functions like the Gauss hypergeometric function. These systems are defined by a matrix of integers A and a parameter vector, and their solutions can be expressed as formal power series. The central question is how well a particular method of constructing solutions, called intrinsic perturbation, covers the full space of solutions to these systems.
The intrinsic perturbation method produces specific series solutions by slightly tweaking certain parameters in a systematic way. The authors carefully measure how much of the total solution space these series actually span. They introduce a precise formula for the "codimension," meaning the size of the gap between the span of the intrinsically perturbed series and the entire solution space. A codimension of zero would mean the method captures everything; a large codimension means many solutions are missed. The authors show this gap, measured in a formal algebraic setting, also appears when you look at genuine holomorphic (complex analytic) solutions on a suitable domain.
To illustrate how large this gap can be in practice, the authors study a specific family of examples where the matrix A has a particular structure controlled by an integer q. For these cases, they prove that the codimension grows roughly like three-quarters of q squared as q increases. This means that as the complexity of the system grows, the intrinsic perturbation method misses an increasingly large portion of the solution space. The result gives a concrete lower bound showing the limitations of intrinsic perturbation as a tool for generating all solutions to A-hypergeometric systems.