How does the topological space of science emerge? Inspired by the concept of maps of science, i.e. mapping scientific topics to a scientific space, we ask which topological structure a dynamical process of authors collaborating and publishing papers can generate. We propose a dynamical process where papers as well as new groups receive topical positions embedded in a two-dimensional euclidean space. The precise position of new papers depends on previous topics of the respective authors and is chosen randomly in a surrounding neighborhood including novelty and interdisciplinarity. Depending on parameters, the spatial structure resembles a simple Gaussian distribution, or spatial clusters of side-topics are observed. We quantify the time-evolution of the spatial structure and discuss the influence of inhomogenities.