Source and evolution of fluids in the Haidewula uranium deposit, Northwest China: evidence from elemental and isotopic compositions of pyrite

• Elemental and isotopic data of pyrite reveal fluid source and characteristics. • A data-driven framework created for predictive uranium exploration. • Low Ni, high Mo, and high Pb serve as key geochemical indicators for hydrothermal uranium deposits. The Haidewula uranium deposit, the first volcanic-type discovery on the Qinghai-Tibet Plateau, provides new insights into ore-forming processes, fluid evolution, and source materials. This study integrates petrographic observations with in situ LA-ICP-MS trace element and sulfur isotope (δ 34 S) analyses of pyrite to constrain fluid evolution and deposit genesis. Three pyrite generations are distinguished: pre-ore (Py1), main ore-stage (Py2), and late ore-stage (Py3). The geochemical signatures of these pyrite generations are consistent with hydrothermal origins. The enrichment of U, Mo, and W in ore-stage pyrite, especially Py2, indicates derivation from Silurian intermediate–acidic volcanic rocks. Sulfur isotope data reveal a complex fluid evolution. Sulfur in Py1 is derived from Paleozoic sedimentary strata, Py2 records a mixture of basin-derived and mantle-derived fluids, and Py3 reflects a transition to higher oxygen fugacity. Trace element patterns, including enrichment of As, Tl, Pb, and Bi accompanied by Se depletion, indicate medium- to low-temperature conditions and progressive cooling. Uranium deposition was facilitated by wall-rock alteration and fluid decompression, destabilizing uranyl complexes. Machine learning models classify deposit types effectively, with Random Forest showing the highest accuracy. Elevated Cd and Mo concentrations in pyrite are identified as robust geochemical fingerprints of volcanic-type uranium systems, providing valuable criteria for exploration. Collectively, these findings not only clarify the genetic mechanisms of Haidewula but also establish a methodological framework that advances predictive models for volcanic-type uranium exploration worldwide.