Title:Communicability and multipartite structures in complex networks at negative absolute temperatures

Abstract: We here present a method of clearly identifying multi-partite subgraphs in a network. The method is based on a recently introduced concept of the communicability, which very clearly identifies communities in a complex network. We here show that, while the communicability at a positive temperature is useful in identifying communities, the communicability at a negative temperature is useful in idenfitying multi-partitite subgraphs; the latter quantity between two nodes is positive when the two nodes belong to the same subgraph and is negative when not. The method is able to discover `almost' multi-partite structures, where inter-community connections vastly outweigh intracommunity connections. We illustrate the relevance of this work to real-life food web and protein-protein interaction networks.