Triangle-free subgraphs at the triangle-free process

We consider the triangle-free process: given an integer n, start by taking a uniformly random ordering of the edges of the complete n-vertex graph Kn. Then, traverse the ordered edges and add each traversed edge to an (initially empty) evolving graph unless its addition creates a triangle. We study the evolving graph at around the time where Θ(n) edges have been traversed for any fixed ε ∈ (0, 10). At that time and for any fixed triangle-free graph F , we give an asymptotically tight estimation of the expected number of copies of F in the evolving graph. For F that is balanced and have density smaller than 2 (e.g., for F that is a cycle of length at least 4), our argument also gives a tight concentration result for the number of copies of F in the evolving graph. Our analysis combines Spencer’s original branching process approach for analysing the triangle-free process and the semi-random method.