Patterns of Collaboration in Co-authorship Networks in Chemistry - Mesoscopic Analysis and Interpretation

We are investigating the manner in which different scientific fields differ in their communication cultures. Our research methodology combines quantitative (graph-theoretic) and qualitative (ethnographic) techniques, which mutually inform each other. In this paper, we present the results of a case study on the identification and classification of scientific collaborations based on structural patterns in co-authorship networks. We investigate three co-authorship networks in specialised subfields of organic chemistry and physical chemistry using an information-theoretic clustering algorithm by Rosvall et al. to extract the modular structure of the networks and thereby investigate the mesoscopic structures of these networks. The clusters found mostly correspond to hierarchically organised research groups or multi-centre research networks. Based on the co-author links between these clusters we analyse group and network interactions, using interviews with participants from the fields in question to inform our interpretation. Structurally one can distinguish two broad classes of linking patterns between clusters: 'exclusive' connections where a single node (author) connects two clusters, and 'manyto-many' connections where substantial fractions of the nodes of both clusters are involved. Within these broad classes one can find further subclasses that correspond to typical collaboration and mobility events.