Research institutions provide the infrastructure for scientific discovery, yet their role in the production of knowledge is not well characterized. To address this gap, we analyze interactions of researchers within and between institutions from millions of scientific papers. Our analysis reveals that collaborations densify as each institution grows, but at different rates (heterogeneous densification). We also find that the number of institutions scales with the number of researchers as a power law (Heaps’ law) and institution sizes approximate Zipf’s law. These patterns can be reproduced by a simple model in which researchers are preferentially hired by large institutions, while new institutions complimentarily generate more new institutions. Finally, new researchers form triadic closures with collaborators. This model reveals an economy of scale in research: larger institutions grow faster and amplify collaborations. Our work deepens the understanding of emergent behavior in research institutions and their role in facilitating collaborations.

研究机构为科学发现提供基础设施，但它们在知识生产中的作用并没有得到很好的描述。为了弥补这一差距，我们从数百万篇科学论文中分析了机构内部和机构之间研究人员的互动。我们的分析表明，随着每个机构的发展，合作变得更加密集，但速度不同（异构密集化）。我们还发现机构的数量与研究人员的数量呈幂律关系（Heaps 定律），机构规模近似于 Zipf 定律。这些模式可以通过一个简单的模型来复制，其中研究人员优先被大型机构聘用，而新机构则互补地产生更多的新机构。最后，新的研究人员与合作者形成三元闭包。这个模型揭示了研究的规模经济：更大的机构发展得更快并扩大合作。我们的工作加深了对研究机构中紧急行为及其在促进合作中的作用的理解。