As the critical elements of scientific literature, mathematical formulas contribute significantly to the effectiveness of literature retrieval. Existing formula-based retrieval methods of scientific literature primarily focus on the number of matching symbols between formulas and often neglect the semantic relevance which leads to suboptimal retrieval outcomes. Additionally, existing literature retrieval methods frequently ignore the ontological relationships between literature, resulting in potential semantic information loss. To address these challenges, we propose a two-stage scientific literature retrieval model named SC-HNE (Symbol Contribution-Heterogeneous Network Enhancement). First, we introduce a Symbol Contribution Aware (SCA) module that calculates the contribution of each symbol based on its hierarchical position in the Operator Tree (OPT), and the symbols with the higher contributions will be paid more attention during the process of formula matching. Second, a heterogeneous network for scientific literature is constructed which uses Heterformer to enrich literature node embeddings with network-aware information and enhance literature representation. Finally, the similarity scores from two modules are combined to rank the top-K scientific literature. Experiments on the DBLP-Citation-network dataset demonstrated the effectiveness of the proposed method through MAP and NDCG metrics.