The increasing importance of industry-university-research institute (IUR) collaborations in knowledge commercialization is widely acknowledged. Research and development (R&D) activities between firms have a positive spillover effect, enhancing innovation capabilities, especially within the same knowledge domain. This effect is amplified through strategic alliances with universities and research institutes. Various collaboration models exist globally (e.g., I/UCRC in the US, enterprise-focused R&D in Germany, commissioned research in Japan), with China experiencing rapid growth in R&D alliance networks. This paper aims to investigate the formation and evolution of R&D alliance networks in China's IUR collaboration, identifying key players in technology and knowledge dissemination across different time periods. Three networks—collaboration, knowledge, and inter-organizational technology networks—will be constructed and analyzed using two-mode network analysis to examine their overall structure, individual characteristics, and temporal evolution.

Existing research utilizes network analysis to study collaborations, focusing on co-author relationships in academic papers or knowledge/technology exchange between organizations in joint patent applications. However, a unified definition of network formats is lacking, due to the complexity of R&D alliance networks with multi-level links. Three main network types are identified: collaboration networks (emphasizing cooperation relationships), knowledge networks (emphasizing knowledge flow), and inter-organizational technology networks (linking organizations and technologies). Two-mode network analysis, recognizing organizations' participation in multiple networks, offers a deeper understanding. While prior work has applied two-mode analysis at the country level, this study takes a micro-level, organizational perspective, integrating inter-organizational and technology networks to fill a research gap.

The study focuses on China's strategic emerging industries (energy-saving, new energy vehicles, high-end equipment manufacturing, new energy, new materials, new generation information technology, and bio-industries). Data were collected from the China State Intellectual Property Office (SIPO) database, using a keyword-based search strategy to identify relevant patents (1995-2018). A total of 26,704 patents were initially identified, with 11,763 qualifying patents selected after careful review and filtering. Four-digit International Patent Classification (IPC) codes were used as proxies for knowledge elements. Three types of networks were constructed: collaboration networks (based on joint patent assignees), knowledge networks (based on shared IPC codes among assignees), and inter-organizational technology networks (linking organizations and patents). A five-year rolling window analysis was employed using Sci2 Tool software. Network topological analysis (network scale, density, centrality, average distance, cohesion index, clustering coefficient) and ego network analysis (degree centrality, betweenness centrality, structural hole) were conducted to analyze the overall network structure and individual node characteristics across different time periods.

The study's analysis of the collaboration network revealed a shift from a less interconnected structure in earlier periods (1995-2003) to a denser, more centralized structure in later periods (2004-2017). Table 1 presents the topological indicators showing these changes (number of edges, density, clustering coefficient, average shortest path, centralization index). Figures 5a-e visually depict the evolution of the collaboration network's topology across different periods (2007-2011, 2008-2012, 2009-2013, 2010-2014, 2011-2015), highlighting the emergence of core organizations with high centrality. The analysis identified specific organizations with high centrality in each period, including universities, research institutes, and enterprises (e.g., Siemens (China) Co., Ltd., Central South University of Forestry and Technology, State Grid Corporation of China, etc.). The knowledge network analysis (Table 2, Figures 6a-e) showed a similar trend, progressing from a less connected network to a more complex, yet less dense, network over time. This reflects the increasing diversification of knowledge elements. The study further identified leading knowledge elements (IPC codes) across different time periods, showcasing the evolution of core technologies within the IUR collaborations. Finally, the inter-organizational technology network analysis (Table 3, Figures 7a-e) showed a consistent increase in the number of edges and average degree over time, indicating increasing complexity. Again, specific organizations and core technologies (e.g., State Grid Corporation of China, H02J3, G01R31) were identified as prominent nodes within the network. The evolution of the networks suggests three phases: formation (2000-2010), growth (2011-2014), and maturity (2014-2018), each characterized by specific collaboration modes (U-R, I-U, I-R, I-U-R) and dominant organizations (Tables 4, 5, 6, 7). The study revealed that there is no single permanent "superstar" consistently dominating the network but a change in the role of universities and corporations in different periods. Government policies such as the "2011 plan" influenced the network structure and the role of universities.

The findings address the research question by revealing the dynamic nature of R&D alliance networks in China's IUR collaboration. The identification of distinct collaboration modes and key players across different periods highlights the importance of understanding the network's evolutionary trajectory. The absence of permanent "superstars" suggests that collaboration success depends on the adaptability of organizations to changing network structures and technological landscapes. The results are relevant to the field by providing a nuanced understanding of IUR collaboration in a developing country context, highlighting the role of government policy and the interplay between different types of organizations. The study provides insights into the factors driving network evolution and the strategies that organizations can employ to enhance their position and success within these networks.

This study makes significant contributions by employing two-mode network analysis to unravel the dynamics of R&D alliance networks in China's IUR collaboration. The identification of distinct evolutionary phases and key players across these phases provides valuable insights for policy-makers and organizations involved in such collaborations. Future research could explore the impact of other factors (e.g., regional context, firm characteristics) on network dynamics, expand the analysis to multi-mode networks, and refine the network analysis methods. Further research could also compare the findings with similar analyses in developed countries to better understand the differences and similarities in network structures and dynamics.

The study has limitations. The data are from the SIPO database, potentially excluding patents from other sources. The analysis focuses on a two-mode network, neglecting the influence of other factors (e.g., regional context, internal firm structures). Further research could incorporate additional network measures (e.g., network diameter, efficiency, block model) and benchmark the empirical network against simulated networks to better understand network evolution.