Overview
Extremal graph theory asks how far a graph parameter can be pushed under a structural restriction. A central example maximizes the number of edges without creating a particular subgraph. A successful solution supplies both a construction attaining the proposed value and an argument excluding every larger value.
Prerequisites
Use finite simple graphs, degrees, subgraphs, complete graphs, and basic inequalities. Distinguish an ordinary subgraph from an induced subgraph: extra edges among selected vertices matter only for the induced version.
Definition and conventions
This entry focuses on forbidden ordinary subgraphs. Fix a finite simple graph containing at least one edge. For an integer , define to be the largest edge count among -vertex simple graphs containing no subgraph isomorphic to .
Thus -free means free of ordinary copies, not merely induced copies. The maximum exists because there are finitely many edge sets on a fixed labeled vertex set. If , then . Other extremal problems optimize minimum degree, clique counts, or parameters of set systems; they require their own specifications.
Intuition
Adding edges increases density but can create a forbidden pattern. Extremal constructions distribute edges so that many pairs are connected while certain configurations remain impossible. Upper bounds exploit the same obstruction locally and then sum it across the whole graph.
Examples and non-examples
A complete bipartite graph with parts of sizes three and four has seven vertices and twelve edges. It has no triangle, because any three vertices include two in one part, and those two are nonadjacent. Mantel’s theorem says twelve is the maximum possible: .
For a convention check, the complete graph on four vertices contains an ordinary three-vertex path: select three vertices and retain two of their edges. It has no induced three-vertex path, since all three selected edges remain in an induced subgraph. Therefore a prohibition of ordinary paths cannot be replaced by a prohibition of induced paths.
Important properties
Mantel’s bound has a useful degree proof. Let a triangle-free graph have vertices and edges. For each edge , its endpoint neighborhoods are disjoint, giving . Summing over edges yields .
Cauchy–Schwarz and the degree sum identity give . Combining and dividing by gives . Since is integral, take the floor. The cases and satisfy the bound directly.
For larger forbidden cliques, Turán’s theorem identifies the complete -partite graph with part sizes differing by at most one as an extremal -free graph, for integers . Empty parts are allowed when .
Connections and further reading
Ramsey theory uses forbidden patterns in complementary colors. Random constructions often provide lower bounds when balanced partitions no longer suffice. Stability asks whether graphs near the extremal edge count resemble an extremal construction. Zhao, Chapter 1, §§1.1–1.2 develops the triangle and clique cases before more difficult forbidden graphs.