On the Complexity of General Graph Factor Problems

For arbitrary graphs G and H, a G-factor of H is a spanning subgraph of H composed of disjoint copies of G. G-factors are natural generalizations of l-factors (or perfect matchings), in which G replaces the complete graph on two vertices. Our results show that the perfect matching problem is essentially the only instance of the G-factor problem that is likely to admit a polynomial time bounded solution. Specifically, if G has any component with three or more vertices then the existence question for G-factors is NP-complete. (In all other cases the question can be resolved in polynomial time.)

The notion of a G-factor is further generalized by replacing G by an arbitrary family of graphs. This generalization forms the foundation for an extension of the traditional theory of matching. This theory, whose details will be developed elsewhere, includes, in addition to further NP-completeness results, new polynomial algorithms and simple duality results. Some indication of the nature and scope of this theory are presented here.