Graph patterns on the web are characteristic substructures of web graphs. Graph patterns have played important roles in web analysis and information retrieval. However, temporal characteristics of the web have not been estimated sufficiently in previously proposed graph patterns. We first propose time graph patterns, and then, we mine the patterns representing topics that are discussed extensively on the web. Conducting experiments using those mined patterns, we finally ascertain that time graph pattern opens a novel way for information retrieval and web analysis reflecting temporal characteristics.