Forbidden Induced Subgraphs and the ĆoĆ-Tarski Theorem
Let đ be a class of finite and infinite graphs that is closed under induced subgraphs. The well-known ĆoĆ-Tarski Theorem from classical model theory implies that đ is definable in first-order logic (FO) by a sentence Ï if and only if đ has a finite set of forbidden induced finite subgraphs. It provides a powerful tool to show nontrivial characterizations of graphs of small vertex cover, of bounded tree-depth, of bounded shrub-depth, etc. in terms of forbidden induced finite subgraphs. Furthermore, by the Completeness Theorem, we can compute from Ï the corresponding forbidden induced subgraphs. We show that this machinery fails on finite graphs. - There is a class đ of finite graphs which is definable in FO and closed under induced subgraphs but has no finite set of forbidden induced subgraphs. - Even if we only consider classes đ of finite graphs which can be characterized by a finite set of forbidden induced subgraphs, such a characterization cannot be computed from an FO-sentence Ï, which defines đ, and the size of the characterization cannot be bounded by f(|Ï|) for any computable function f. Besides their importance in graph theory, the above results also significantly strengthen similar known results for arbitrary structures.
READ FULL TEXT