Testing first-order properties for subclasses of sparse graphs

We present a linear-time algorithm for deciding first-order (FO) properties in classes of graphs with bounded expansion, a notion recently introduced by Nesetril and Ossona de Mendez. This generalizes several results from the literature, because many natural classes of graphs have bounded expansion: graphs of bounded tree-width, all proper minor-closed classes of graphs, graphs of bounded degree, graphs with no subgraph isomorphic to a subdivision of a fixed graph, and graphs that can be drawn in a fixed surface in such a way that each edge crosses at most a constant number of other edges. We deduce that there is an almost linear-time algorithm for deciding FO properties in classes of graphs with locally bounded expansion. More generally, we design a dynamic data structure for graphs belonging to a fixed class of graphs of bounded expansion. After a linear-time initialization the data structure allows us to test an FO property in constant time, and the data structure can be updated in constant time after addition/deletion of an edge, provided the list of possible edges to be added is known in advance and their simultaneous addition results in a graph in the class. All our results also hold for relational structures and are based on the seminal result of Nesetril and Ossona de Mendez on the existence of low tree-depth colorings.

💡 Research Summary

The paper “Testing first‑order properties for subclasses of sparse graphs” establishes a powerful algorithmic meta‑theorem for a broad family of sparse graphs known as classes of bounded expansion. Bounded expansion, introduced by Nešetřil and Ossona de Mendez, captures many familiar sparse graph families—graphs of bounded tree‑width, all proper minor‑closed classes, bounded‑degree graphs, graphs excluding a fixed subdivision, and graphs that can be drawn on a fixed surface with a constant number of crossings per edge. The authors show that for any class G with bounded expansion and any first‑order (FO) property Π, there exists a linear‑time algorithm that decides whether a given graph from G satisfies Π.

The technical core relies on the existence of low‑tree‑depth colourings for bounded‑expansion graphs. Such a colouring uses a constant number of colours (depending only on the class) and guarantees that each colour class induces a subgraph of bounded tree‑depth, i.e., a shallow rooted forest. By converting an FO sentence into prenex normal form and assigning each quantified variable to a colour, the search space for variable assignments collapses from the whole vertex set to a constant‑size neighbourhood within the corresponding shallow forest. Consequently, evaluating the quantifier‑free part of the formula reduces to constant‑time table look‑ups on each colour class, and the overall evaluation runs in O(|V(G)| + |φ|) time, where φ is the FO sentence.

Beyond the static setting, the authors design a dynamic data structure that, after O(n) preprocessing (n = |V(G)|), supports: (i) insertion or deletion of a tuple in any relational symbol of the guarded structure in O(1) time (provided the underlying graph remains in the fixed bounded‑expansion class), and (ii) answering whether the current structure satisfies a Σ₁‑FO sentence with at most d₀ variables in O(|φ|) time, also returning a satisfying assignment if one exists. The data structure stores, for each colour class, the shallow forest representation and pre‑computed tables for all possible atomic relations; updates affect only the colour class containing the modified tuple, preserving constant‑time performance.

The paper further extends the results to classes of locally bounded expansion. A class has locally bounded expansion if, for every radius d, the d‑neighbourhood of any vertex satisfies a bounded‑expansion density bound. By applying low‑tree‑depth colourings locally and stitching the results together, the authors obtain an “almost linear‑time” algorithm: for any ε > 0, the FO model‑checking problem can be solved in O(n^{1+ε}) time on such classes.

The authors also discuss the relationship with nowhere‑dense graph classes, which are a strict superset of locally bounded‑expansion families. While their techniques do not directly yield linear‑time FO model‑checking on nowhere‑dense graphs, they present a dynamic data structure for these classes that achieves O(n^{1+ε}) preprocessing and O(n^{ε}) update time for any fixed ε > 0, while still answering Σ₁‑queries in O(|φ|) time. Moreover, they prove a hardness result: if Σ₁‑model‑checking on a non‑nowhere‑dense class were fixed‑parameter tractable (parameterised by the formula size), then FPT would equal W