Local-Order-Invariant Logic on Classes of Bounded Degree
Derek Aoki
TL;DR
The paper investigates order-invariant, local-order-invariant, and epsilon-invariant logics within finite model theory, addressing how locality notions constrain their expressive power. It develops a Hanf locality–based approach, introducing R-presentation schemes and locality notions to analyze invariant logics. The authors prove that local-order-invariant logic collapses to first-order logic on classes of bounded degree, and show similar collapses for epsilon-invariant logic in the same setting, while also providing upper bounds and a decomposition method to extend these results to broader finite structures. Additionally, they demonstrate non-collapse results for order- and local-order-invariant logics on general finite or locally finite structures, and explore the relationships between these logics under bounded-diameter conditions, suggesting directions for future upper-bound techniques.
Abstract
Local-order-invariant (first-order) logic is an extension of first-order logic where formulae have access to a ternary local order relation on the Gaifman graph, provided that the truth value does not depend on the specific order relation chosen. Weinstein asked a number of questions about the expressive power of order-invariant and local-order-invariant logics on classes of finite structures of bounded degree, classes of finite structures in general, and classes of locally finite structures. We answer four of his five questions, including showing that local-order-invariant logic collapses to first-order logic on classes of bounded degree. We also investigate epsilon-invariant logic. We show that epsilon-invariant logic collapses to first-order logic on classes of bounded degree by containing it in local-order-invariant logic in this setting, and we give an upper bound for epsilon-invariant logic in terms of local-order-invariant logic on general finite structures. Finally, in the process of proving these theorems, we demonstrate some principles which suggest further directions for showing upper bounds on invariant logics, including an upper bound on epsilon-invariant logic in general.