The Yamansu deposit, which is hosted in the volcanic-sedimentary sequence of the Carboniferous Yamansu Formation in Eastern Tianshan, NW China, contains many skarns, and the orebodies occur in the ore district in stratoidal, banded or lenticular forms. Four alteration stages, namely, albite–tourmaline–apatite–Grt1 (Stage I), K-feldspar–Grt2 (Stage II), magnetite–chlorite–epidote (Stage III), and quartz–calcite–axinite–Grt3 (Stage IV), are distinguished in the Yamansu deposit. The mineral geochemistry associated with each different stage is presented to provide a better understanding of the corresponding metallogenic processes. The ore-forming fluid in Stage I was derived from a magmatic–hydrothermal source and formed at high temperatures with many volatiles. This ore-forming fluid, which contained considerable metallogenic materials during the early stage, likely experienced diffusive metasomatism in a closed system with low water/rock (W/R) ratios. Mineral geochemical analyses show that the Fe content gradually increases from Stage I to Stage II, indicating that accumulated ore-forming materials were available during changes in the physicochemical conditions from a reducing environment with neutral pH to oxidizing conditions with mildly acidic pH. During the main metallogenic stage (Stage III), mineral assemblages reflect moderate- to high-temperature conditions, and the ore-forming fluid was created and destroyed periodically; the magnetite ores were deposited in a fluctuating fluid system. The multilayered orebodies, multigenerational garnets, and minerals with oscillatory zoning indicate that the ore-forming fluid may have developed periodic fluctuations, and this special multistage fluctuation of the hydrothermal fluid in the Yamansu deposit was the key factor controlling the multiple extraction, enrichment and precipitation of metallogenic materials.

The Yamansu iron deposit, hosted in submarine volcanic rocks, is located in the Aqishan–Yamansu Fe–Cu metallogenic belt of eastern Tianshan, Xinjiang. New pyrite Re–Os and pyroxene diorite intrusion ages and the chemistry of magnetite constrain the origin of the deposit. The Yamansu orebodies display banded or lenticular forms and sharp contacts with marble and garnet skarn. Pyrite associated with magnetite from hydrothermal massive ores has a Re–Os isochron age of 320.3 ± 9.1 Ma, which is consistent with the Carboniferous age of volcanic rocks. However, the pyroxene diorite age (252.1 ± 3.7 Ma) is obviously younger than the mineralization age, implying no temporal relationship between mineralization and intrusion. Magnetite samples from different ores in the Yamansu deposit are suggested to have a magmatic–hydrothermal origin. The magmatic primary magnetite is rich in TiO₂ and poor in SiO₂, MgO, and CaO, similar to magnetite from Kiruna type deposits. The hydrothermal magnetite is rich in FeO, CaO, and SiO₂ but poor in TiO₂ and shows oscillatory zoning. Moreover, magmatic and hydrothermal magnetite samples show different characteristic normalized rare earth element and trace element patterns. A combination of ore fabrics, discrimination diagrams, and normalized patterns for magnetite samples can reflect the magmatic–hydrothermal process. The Yamansu iron deposit is spatially and temporally associated with Carboniferous volcanism, and we propose that the magmatic magnetite ores were derived from iron-rich melt through melt immiscibility. The residual iron-rich magma ascended and erupted along with hydrothermal activity; then moderate to large amounts of hydrothermal massive, banded, or disseminated magnetite formed with skarns.

This paper documents the zircon U–Pb ages, whole-rock geochemistry, and Sr–Nd–Pb isotopes of the Mesozoic granites in the central part of the North Qilian Orogenic Belt to provide information on the tectonic evolution and crustal accretion process of the Qilian Orogenic Belt. Zircon U–Pb dating yields an age of 215.3 ± 3.1 Ma, indicating that the Beidaban monzogranites formed from Late Triassic. They are characterized by high contents of SiO₂, Al₂O₃, and K₂O; are slightly peraluminous (A/CNK = 1.08–1.15); and have mineralogical assemblages of primary biotite and ilmenite, illustrating that they are shoshonitic and peraluminous S-type granite. The Beidaban monzogranites have initial (⁸⁷Sr/⁸⁶Sr)_i values ranging from 0.71456 to 0.71867 and εNd(t) values ranging from –12.9 to –8.5 with two-stage Nd model ages of 1.69–2.04 Ga, suggesting that they originated from partial melting of the Paleo-Mesoproterozoic (Longshoushan Group) continental crustal metasedimentary rocks. Initial Pb isotopic compositions (²⁰⁶Pb/²⁰⁴Pb = 19.44–21.80; ²⁰⁷Pb/²⁰⁴Pb = 15.76–15.89; ²⁰⁸Pb/²⁰⁴Pb = 39.62–41.26) and geochemical features such as high Th/Ta (9.3–67.4, 37.4 on average) and Rb/Nb (12.5–17.1) are consistent with recycled crustal components. Combined with previous geochronological and geochemical data, we suggest that the Mesozoic granites probably formed in a post-collisional tectonic setting and that the North Qilian Orogen Belt experienced comprehensive intracontinental orogenesis after the closure of the Qilian ocean.

The Kalatag Cu–polymetallic district in the Eastern Tianshan Orogen, NW China, is characterized by complex tectono-magmatic evolution and diverse mineralization styles, including porphyry-skarn, hydrothermal-vein, and volcanogenic massive sulfide (VMS) systems. However, only nine verified mineral occurrences were available for calibration of the present regional prospectivity model. This small occurrence dataset, together with the heterogeneous spatial distribution of geological and geochemical indicators, poses a challenge for regional prospectivity mapping. In this study, a Maximum Entropy (MaxEnt) model was developed to evaluate Cu–polymetallic prospectivity using presence-only mineral occurrence data. Seven rock-debris geochemical predictors, including Cu, Zn, Au, Mo, Ag, Ni, and Sb, were selected based on statistical characteristics and metallogenic relevance, while two geological predictors representing proximity to faults and intrusions were derived from regional geological maps. Nine-fold leave-one-out cross-validation (LOOCV) was used to evaluate out-of-sample discrimination and model sensitivity to regularization. The interpretation-oriented RM = 0.5 configuration yielded a mean LOOCV test AUC of 0.665, whereas the final model fitted using all nine verified occurrences achieved a training AUC of 0.850. High-prospectivity zones occupied 12.20% of the study area and contained 88.89% of known mineral occurrences. Jackknife analysis and response curves indicate that distance to intrusions, distance to faults, and Au anomalies provide the strongest independent spatial information within the predictor system. Prospectivity is generally enhanced near fault zones and intrusive contacts, highlighting the importance of tectono-magmatic architecture in controlling the spatial distribution of Cu–polymetallic mineralization. The resulting prospectivity map identifies regional high-prospectivity zones where favorable geological and geochemical conditions coincide, providing a quantitative basis for exploration prioritization. The study indicates that MaxEnt is an effective approach for integrating multisource geoscientific data under small-sample and presence-only conditions, while emphasizing that model outputs require further validation through geological, geochemical, geophysical, and drilling investigations.