Abstract:     The exploration of breakthrough innovation is vital for enterprises to overcome the inertia created by reliance on technological import and imitation. This effort is essential for China to become self-reliant in advanced science and technology sectors. This study, based on the knowledge-based view and utilizing a knowledge recombination perspective, integrated creatively multiple networks from inventor-knowledge bipartite graphs into a singular framework. It explored how the characteristics of these bipartite networks influence breakthrough innovation, with an empirical analysis conducted using negative binomial regression on data from publicly listed semiconductor firms from 2000 to 2021. The results of the negative binomial regression revealed that inventor knowledge diversity positively affects breakthrough innovation while knowledge uniqueness negatively affects it. Enterprises′ knowledge recombination capability plays a partial mediating role between inventor-knowledge two-mode network characteristics and breakthrough innovation. The structural hole efficiency of the inventor network negatively moderates the positive effect of inventor knowledge diversity, but positively moderates the mediating role of knowledge recombination capability. The structural hole efficiency in the knowledge network lessens both the negative impact of knowledge uniqueness on recombination capability and its mediation effect. The findings have revealed multiple network embedding mechanisms for breakthrough innovation, and will provide valuable insights to help enterprises enhance their breakthrough innovation.

Key words: breakthrough innovation, inventor-knowledge two-mode network, knowledge recombination and creation capability, structural hole efficiency