A planar graph is a graph whose vertices can be represented by points in the Euclidean plane, and whose edges can be represented by simple curves in the same plane connecting the points representing their endpoints, such that no two curves intersect except at a common endpoint. Planar graphs are often drawn with straight line segments representing their edges, but by Fáry's theorem this makes no difference to their graph-theoretic characterization. A subdivision of a graph is a graph formed by subdividing its edges into paths of one or more edges. Kuratowski's theorem states that a finite graph G is planar, if it is not possible to subdivide the edges of K5 or K3,3, and then possibly add additional edges and vertices, to form a graph isomorphic to G. Equivalently, a finite graph is planar if and only if it does not contain a subgraph that is homeomorphic to K5 or K3,3.
Kuratowski subgraphs
If G is a graph that contains a subgraph H that is a subdivision of K5 or K3,3, then H is known as a Kuratowski subgraph of G. With this notation, Kuratowski's theorem can be expressed succinctly: a graph is planar if and only if it does not have a Kuratowski subgraph. The two graphs K5 and K3,3 are nonplanar, as may be shown either by a case analysis or an argument involving Euler's formula. Additionally, subdividing a graph cannot turn a nonplanar graph into a planar graph: if a subdivision of a graph G has a planar drawing, the paths of the subdivision form curves that may be used to represent the edges of G itself. Therefore, a graph that contains a Kuratowski subgraph cannot be planar. The more difficult direction in proving Kuratowski's theorem is to show that, if a graph is nonplanar, it must contain a Kuratowski subgraph.
A Kuratowski subgraph of a nonplanar graph can be found in linear time, as measured by the size of the input graph. This allows the correctness of a planarity testing algorithm to be verified for nonplanar inputs, as it is straightforward to test whether a given subgraph is or is not a Kuratowski subgraph. Usually, non-planar graphs contain a large number of Kuratowski-subgraphs. The extraction of these subgraphs is needed, e.g., in branch and cut algorithms for crossing minimization. It is possible to extract a large number of Kuratowski subgraphs in time dependent on their total size.
History
published his theorem in 1930. The theorem was independently proved by Orrin Frink and Paul Smith, also in 1930, but their proof was never published. The special case of cubic planar graphs was also independently proved by Karl Menger in 1930. Since then, several new proofs of the theorem have been discovered. In the Soviet Union, Kuratowski's theorem was known as either the Pontryagin–Kuratowski theorem or the Kuratowski–Pontryagin theorem, as the theorem was reportedly proved independently by Lev Pontryagin around 1927. However, as Pontryagin never published his proof, this usage has not spread to other places.
Related results
A closely related result, Wagner's theorem, characterizes the planar graphs by their minors in terms of the same two forbidden graphs K5 and K3,3. Every Kuratowski subgraph is a special case of a minor of the same type, and while the reverse is not true, it is not difficult to find a Kuratowski subgraph from one of these two forbidden minors; therefore, these two theorems are equivalent. An extension is the Robertson-Seymour theorem.
