[S-CG46] New Insights of Fluid-Rock Interactions: From Surface to Deep Subduction Zone

[SCG46-P13] A machine learning-based method for discriminating the ore-forming process in magmatic-hydrothermal systems

*Denghui Zhu¹, Shen Gao², Jiajie Wang¹, Noriyoshi Tsuchiya¹ (1.Graduate School of Environmental Studies, Tohoku University, 2.China University of Geosciences, Beijing)

Keywords:quartz, magmatic-hydrothermal system, machine learning, identification

Quartz is a widespread gangue mineral in magmatic-hydrothermal systems. Trace elements in quartz record physical and chemical evolutions of quartz growth, providing essential clues about the formation processes of ore deposits in magmatic-hydrothermal systems. Previous studies utilize quartz chemistry such as Al verse Ti biplot and Ti-Al-Ge ternary plot to identify deposit types in magmatic-hydrothermal systems. However, two or three-dimensional data cannot capture all the features of trace elements, leading to a lower accuracy on the identification of deposit types. Considering the promising advantages of machine learning technique for processing large-sized and high-dimensional data, this research aims to propose a new machine learning-based method to analyze trace element data (Li Be B Na Mg Al K Ca Ti Cr Mn Fe Cu Zn Ga Ge As Rb Sr Y Mo Ag Sn Sb Cs Ba Pb Th U) in quartz of magmatic-hydrothermal system to provide a more efficient and precise interpretation of the origin and deposit types of quartz in magmatic-hydrothermal systems.

In this research, we analyzed the quartz trace element data (N = ~2000) of various deposit types and paragenetic stages in porphyry-epithermal system and granite-related uranium system from China by Random Forest (RF) and Uniform Manifold Approximation and Projection (UMAP). Unlike the conventional method, the result showed that utilizing Sb As Be Ge Al can better classify the quartz into porphyry-epithermal system and granite-related uranium system with the F1-score of 93.4%. For granite-related uranium system, utilizing Ti Al Li Sr Ba can further classify the quartz into hydrothermal origin and magmatic origin with the F1-score of 96.6%. For porphyry-epithermal system, utilizing Ti, Sb, Ge, Al, Li, As can classify the quartz into hydrothermal origin, epithermal origin and magmatic origin with the F1-score of 90.7%. Utilizing Sb, Ga, Be, Mg, Li, Sr, Ti can further classify epithermal origin of porphyry-epithermal system into various deposit types with the F1-score of 94.4%.