Concentration and threshold phenomena in random graphs and hypergraphs

ongoing

The central theme of this proposal is the enumeration of large structures obeying a family of local constraints. There are numerous well-studied instances of this general problem in combinatorics, probability theory, and statistical physics. We propose to study a variety of interrelated open problems, spanning extremal and probabilistic combinatorics, Ramsey theory, and large deviation theory, whose unifying theme is the distribution of copies of a given graph in large random graphs. The first part of the proposal revolves around the 'infamous' upper tail problem for subgraph counts in random graphs, which has seen spectacular progress in the last few years. We propose to advance the combinatorial approach introduced in our recent work on a special case of the problem to solve this problem completely and make progress on several related questions. The second part of the proposal discusses the global structure of 'sparse' subsets in hypergraphs. We propose to prove a far-reaching generalisation of the hypergraph container theorem that provides a useful description of the family of all 'sparser-than-average' sets. We plan to use it to resolve the lower tail problem for subgraph counts in the uniform random graph and to prove a sparse analogue of the counting lemma for regular graphs. The third part of the proposal concerns the enumeration of independent sets in hypergraphs with cardinalities below the threshold for appearance of global structure. In order to obtain precise asymptotics for the number of such sets, we propose to extend the cluster expansion method from statistical physics to models with non-pairwise interactions. The fourth part of this proposal deals with thresholds for fundamental extremal and Ramsey properties of random graphs and hypergraphs. We propose to resolve several outstanding open problems in this area that have resisted the recent dramatic advances in the field.