(Submitted on 7 Apr 2009)
Despite the apparent conceptual boundaries of scientific fields, a formal description for their evolution is lacking. Here we describe a novel approach to study the dynamics and evolution of scientific fields using a network-based analysis. We build an idea network consisting of American Physical Society PACS numbers as nodes representing scientific concepts. Two PACS numbers are linked if there exist publications that reference them simultaneously. We locate scientific fields using Cfinder, a community finding algorithm, and describe the time evolution of these fields using a community evolution method over the course of 1985-2006. The communities we identify map to known scientific fields, and their age strongly depends on their size, impact and activity. Our analysis further shows that communities that redefine themselves by merging and creating new groups of ideas tend to have more fitness as measured by the impact per paper, and hence communities with a higher fitness tend to be short-lived. The described approach to quantify the evolution of ideas may be relevant in making predictions about the future of science and how to guide its development.