DENG Yi
1 School of Civil and Resource Engineering, University of Science and Technology Beijing, Beijing 100083, ChinaZHONG Richen
1 School of Civil and Resource Engineering, University of Science and Technology Beijing, Beijing 100083, ChinaLI Dengfeng
2 School of Marine Sciences, Sun Yat-sen University, Zhuhai, Guangdong 519082, China 3 Guangdong Provincial Key Laboratory of Marine Resources and Coastal Engineering, Zhuhai, Guangdong 519082, ChinaLI Yanxia
1 School of Civil and Resource Engineering, University of Science and Technology Beijing, Beijing 100083, ChinaCUI Hao
1 School of Civil and Resource Engineering, University of Science and Technology Beijing, Beijing 100083, ChinaThanks to the editors and two anonymous reviewers for their constructive suggestions and thorough checking of this manuscript. Thanks Chen Huan, Zhong Shuifan, Ling Yifan and Lei Mengdan for providing suggestions on improving the artworks. This work was financially supported by the National Natural Science Foundation of China (Nos. 41872078 and 41502069), the Young Elite Scientists Sponsorship Program by CAST (No. YESS20180143), Fundamental Research Funds for the Central Universities (No. FRF-TP-18-017A3) and the 111 Project of the Ministry of Science and Technology, China (No. BPO719021).
Garnet occurs in a wide range of rock types, from mantle peridotites to granites, from eclogites to skarns. In recent years, garnet LA-ICP-MS (Laser Ablation Inductively Coupled Plasma Mass Spectrometry) U-Pb dating has provided a powerful solution for retrieving the ages of rock formations, but successful dating is often prohibited by the low concentration of U. However, the concentration of U, a trace element of garnet, is unknown prior to the LA-ICP-MS analysis. In this study, we propose that the U concentration in garnet can be predicted by the contents of major and minor elements, which can be quantitatively obtained by EPMA (electron probe microanalysis). Using a supervised machine learning method (neural network), a model is trained to discriminate U-rich (>2 ppm) and U-poor (<2 ppm) garnets, based on EPMA results. Results of cross validation shows that the model has an average accuracy of ~92% and is a powerful tool in detecting datable U-rich garnet. To facilitate the use of the discriminator, it is programmed as a stand-alone Microsoft Excel spreadsheet (HighUGarnet) and users directly paste the molar proportions of garnet end members into it and obtain the discrimination result.
DENG Yi, ZHONG Richen, LI Dengfeng, LI Yanxia, CUI Hao.2022. Hunting the Datable Garnet using the LA-ICP-MS U-Pb Method: Predicting Garnet U Concentration, based on Major and Minor Elements[J]. Acta Geologica Sinica(),96(6):2148-2157
Copy
