The topological classification of geochemical elements is widely used as a reference for regional prospecting prediction. In this study, we analyze the topological correlation structures of 39 representative geochemical elements from the Nanling area of South China by implementing the complex networks theory. The topological correlation structures of geochemical elements have a high clustering coefficient (0.8120-0.8880), but the magnitude of the shortest path (1.2950-2.3600) is small. In combination with the analysis of complex networks characteristics, we report that the topological correlation structures of the geochemical elements in this area have small-world characteristics, which reveals the self-organized criticality. As shown in the topological network, two random elements have some level of associations, which present a specific community feature. Our preliminary result shows that with changing the control parameter (k) of "coarse-graining", the topological correlation structures undergo two critical phase transitions. As the control parameter (k) reaches 0.44, the entire element system evolves into two parts. When the control parameter (k) reaches 0.63, the system forms three "communities". It is worth noting that the three "communities" are basically consistent with the Goldschmidt's geochemical classification of the elements, which are lithophile, siderophile, and chalcophile groups, respectively. In these "communities", we also found that a small level of component units is nested.