Life, death and preferential attachment

Scientific communities are characterized by strong stratification. The highly skewed frequency distribution of citations of published scientific papers suggests a relatively small number of active, cited papers embedded in a sea of inactive and uncited papers. We propose an analytically soluble model which allows for the death of nodes. This model provides an excellent description of the citation distributions for live and dead papers in the SPIRES database. Further, this model suggests a novel and general mechanism for the generation of power law distributions in networks whenever the fraction of active nodes is small.