2.1 Geological setting

The Beiya gold deposit is located in the southern Sanjiang Tethys domain and lies to the east of the Jinshajiang suture (Figure 1), which is situated at the intersection of the western edge of the Yangtze Plate and the southeastern margin of the Himalayan-Tibetan Plateau. The Jinshajiang-Ailao Shan suture was formed at late Permian, which represents the closure of the Jinshajiang-Ailaoshan Ocean (Deng et al., 2015). With the collision of the Indo-Asian continent after Cenozoic, the Ailao Shan-Red River Fault (ASRR) zone was formed along the previous Jinshajiang suture zone (Yin & Harrison, 2000). During 40–30 Ma, due to the typical post-collision setting and the tectonic escape, a crustal-scale extensional movement occurred and a series of alkaline felsic magmas emplaced along the ASRR zone (Yin & Harrison, 2000; Zhou et al., 2017). This resulted in an alkali-rich magmatic belt, which is responsible for the formation of the Beiya gold deposit and numerous Cu-(Mo)-Au deposits along the belt (He et al., 2015; Wang et al., 2015).

The exposed stratigraphy in the Beiya ore district includes the Lower Triassic Qingtianbao Formation, the Middle Triassic Beiya Formation, and Quaternary sedimentary rocks. The Lower Triassic Qingtianbao Formation (175–350 m) comprises sandstone, hornfels greywacke, and mafic volcaniclastic rocks, while the Middle Triassic Beiya Formation (138–531 m) consists of dolomitic, bioclastic, and argillaceous limestone and thin siltstone interbeds. They are covered by N-trending Quaternary sediments (He et al., 2015). Among them, the Beiya Formation is the main host rock for the Beiya skarn Au mineralization (He et al., 2016).

Igneous suites in and around the mining district are the Upper Permian Emeishan Formation, comprising flood basalt, and the locally widespread Cenozoic alkaline porphyries. The Cenozoic alkalic porphyries are thought to have close association with the Beiya Fe-Au skarn mineralization but exposed relatively rare in the district, including quartz–albite porphyry (~65 Ma; Xu et al., 2006; Xu, Cai, Xiao, & Peters, 2007), quartz–syenite porphyry (~37–26 Ma; Fu et al., 2015; Fu, Sun, Zhou, Lin, & Yang, 2016; He et al., 2013; Xu, 2007), and biotite syenite porphyry (3.6–3.8 Ma; He et al., 2015; Liu, Zhou, Chen, Gao, & Huang, 2015; Xu et al., 2006; Xu et al., 2007; Xue et al., 2008). Previous Re-Os dating of molybdenite from Beiya skarn ores yielded Eocene ages of ca. 37.9–34.7 Ma (Fu et al., 2015; He, 2014), which constrains the age of the quartz–sulphide stage, and suggests that the mineralization at Beiya is genetically associated with the quartz syenite porphyry. Overall, the mineralization age is slight younger than the emplacement of the quartz syenite porphyry (Fu et al., 2015, 2016; He, 2014; He et al., 2013).

There are two main fault systems developed in the Beiya district, the N-striking and E-striking fault systems. The N-striking ones are dominant, such as Ma'anshan Fault (a branch of the Jinshajiang-Honghe strike-slip fault system), extending across the western Beiya district, and its secondary faults/fractures are considered to control the emplacement of granitoids in the Wandongshan and Hongnitang ore segments (Li, Wang, et al., 2016). The E-striking fault system was likely post-mineralized and probably dissected or destroyed some orebodies (He, 2014).

The Beiya deposit consists of six ore segments, that is, the Wandongshan, Hongnitang, and Jingouba segments in the western part and Weiganpo, Bijiashan, and Guogaishan segments in the eastern part (Figure 2a). Among them, the Wandongshan segment is the largest, containing 80% of the total resources at Beiya (99 Mt @ 2.61 g/t Au and 170.0 Mt @ 33.34% Fe; Deng et al., 2015).

2.2 Ore deposit geology

Mineralization at Beiya mainly consists of four different types: (a) porphyry Cu-(Mo)-Au mineralization; (b) skarn Au-Cu-Fe mineralization; (c) hydrothermal Pb-Zn-Ag mineralization; and (d) supergene laterite Au-Fe mineralization (Li, Wang, et al., 2016; Zhou et al., 2016). The four distinct mineralization types yield obvious spatial zoning with the porphyry intrusion in the centre and subsequent mineralization occurring outward. Among these major mineralization styles, skarn is the most important mineralization type at Beiya (Zhou et al., 2016).

The Beiya Fe-Au skarn ore bodies occur in and/or along the contact zones between the quartz syenite porphyry intrusion and the carbonate of the Beiya Formation and contain stratiform, vein-like, and laterite ores (Deng et al., 2015; Fu et al., 2016; Zhou et al., 2016). The most representative skarn ore body is the KT52, located in the Wandongshan ore segment (Figure 2b), which has reserves of 87.2 Mt gold ores with an average grade of 2.35 g/t, 90.27 Mt of iron ores (average grade of 34 wt% Fe), 111.8 Mt of copper ores (average grade of 0.34 wt% Cu; Li, Xie, et al., 2016).

Based on field investigation, petrographic observations, and previous studies, the Beiya skarn mineralization can be divided into three major stages: (a) prograde skarn, (b) retrograde skarn, and (c) quartz–sulphide stages (Figure 3). Mineral assemblage in prograde stage is dominated by anhydrous minerals including andradite garnet and diopside. The retrograde stage is predominantly composed by hydrous minerals (e.g., epidote-group minerals, chlorite, amphibole, biotite), magnetite, haematite, titanite, scheelite, fluorite, and feldspar, which largely replaced the garnet and diopside in prograde stage. The quartz–sulphide stage is characterized by abundant pyrite, chalcopyrite, molybdenite, pyrrhotite, sphalerite, galena, arsenopyrite, and bismuth minerals. Bismuth minerals include bismuthinite, cosalite (Figure 4a), tetradymite (Figure 4b), and kitaibelite (Figure 4c), coexisting with pyrite and chalcopyrite or occurring slightly later than them. Quartz–sulphide veins/veinlets locally crosscut massive magnetite, which implies that magnetite formed prior to quartz–sulphide assemblage (Figure 5).

Gold mineralization at Beiya occurs in both late retrograde skarn stage and quartz–sulphide stage (Fu et al., 2017; Zhou et al., 2017). Gold appears as native gold and/or electrum nano- to micro-inclusions (10 nm–10 μm), which can be hosted by pyrite (Figure 4d,e) and magnetite (Figure 4f), or coexist with Bi minerals (Zhou et al., 2017). As an important gold host mineral, magnetite has been suggested to dominate at least 14% of the total Au resources at Beiya (Yunnan Gold and Mineral Resources Crop., pers. commun., 2013).

5.1 Discrimination of magnetite

Magnetite can be crystallized from either silicate, sulphide melts, or hydrothermal fluids, so it is a common accessory mineral in different ore deposit types, for example, porphyry, skarn, IOCG, Kiruna, BIF, and Fe-Ti-V deposits (Dare et al., 2014; Dare, Barnes, & Beaudoin, 2014; Dupuis & Beaudoin, 2011; Nadoll et al., 2012). Combined with the ability to incorporate a number of trace elements into the spinel structure, magnetite has been widely used as a useful petrogenetic indicator and valuable exploration targeting fingerprint (Dare, Barnes, & Beaudoin, 2014; Dare et al., 2014; Dupuis & Beaudoin, 2011; Nadoll et al., 2012, 2014, 2015). Over the past decades, various discriminant diagrams have been established and more grids have been divided based on the large amounts of in situ high-precision data, which have provided good baselines to discriminate various ore deposit types and identify the source of magnetite.

In the Ca + Al + Mn versus Ti + V diagram (based on Dupuis & Beaudoin, 2011; Nadoll et al., 2014), all Beiya magnetites plot in the skarn field (Figure 8a). In the Sn/Ga diagram, the M_E and M_A magnetites plot into the porphyry (hydrothermal) field (Figure 8b), while M_L and M_R magnetites plot into skarn field. In the V/Ti and Ti versus Ni/Cr diagrams, the Beiya data also plot in the hydrothermal field (Figure 8c,d), illustrating that the Beiya magnetites formed from a hydrothermal fluid.

Although many discriminant diagrams have involved Ti as one of the key elements, this element cannot be used alone as a petrogenetic indicator because both high-temperature hydrothermal and magmatic systems have high Ti contents (Duan et al., 2019). Therefore, a multi-element diagram is used to discuss the information from more trace elements at Beiya (Figure 9), in which the trace element concentrations have been normalized to the bulk continental crust (Rudnick & Gao, 2003). In this diagram, all magnetite types basically have the same patterns, that is, they are more similar to the patterns of low-temperature magnetite (blue field), which also constrains the Beiya magnetite a hydrothermal origin.

Notably, dissolution, re-precipitation, and re-crystallization processes during post-ore stage can modify the contents of trace elements in magnetite (Hu et al., 2015; Sun et al., 2017). During these process, Na, Si, Mg, Ca, Mn, Al, Sr, and Cr are mobile, whereas Ti and V show moderately mobile behaviour, and Bi, Co, Sn, and Ga are relatively immobile (Hu et al., 2015; Li et al., 2019). In Beiya, Mn-rich fluids injected along the boundaries of magnetite grains (Figure 10), suggesting that the post-ore fluids could affect the trace elements concentrations in magnetite. Na, K, and Al show discrepancy on the core and rim of one magnetite grain, while the distributions of Pb, Ti, and V are more homogeneous, indicating the former are mobile and the latter are relatively immobile. Thus, a careful microscopic work is needed before further trace elements analysis, and a caution should be paid when using discriminant diagrams (Huang, Zhou, Qi, Gao, & Wang, 2013; Sun et al., 2017; Zhao & Zhou, 2015 and references therein).

5.2 Trace element behaviours in magnetite

The chemical composition of hydrothermal magnetite is variable in different ore deposit types and even in a same type of deposits, for example, trace element contents show discrepancy between different skarn deposits (Nadoll et al., 2015). This is because many factors could govern the composition of hydrothermal magnetite, including the composition of the ore-forming fluids, physicochemical parameters of ore fluids (temperature, pressure, cooling rate, oxygen fugacity, sulphur fugacity, and silica activity), fluid–rock interactions, and the co-precipitation of other minerals (Chen, Zhou, Li, Gao, & Hou, 2015; Dare et al., 2014; Dupuis & Beaudoin, 2011; Huang et al., 2016; Nadoll et al., 2012, 2014). Although there could be more than one factor simultaneously impact on the abundance of trace elements in magnetite, some elements show regularly close relationships with a certain factor, such as Ti, V, Mg, and Mn, Co and Ni concentrations depending more likely on temperature, oxygen fugacity, fluid–rock reaction, and co-precipitation minerals, respectively (Nadoll et al., 2014, 2015).

Titanium is relatively immobile, and its incorporation into magnetite is largely controlled by temperature with greater concentrations always indicating higher temperatures (Dare et al., 2014; Nadoll et al., 2015). In the hydrothermal system, Ti commonly shows a lower concentration (Nadoll et al., 2014). For the Beiya deposit, Ti concentrations within M_E-, M_A-, M_L-, and M_R-type magnetites display a similar content (Figure 8c,d), suggesting these magnetites are precipitated in a relatively narrow range of temperature (Duan et al., 2019; Li et al., 2017). Nevertheless, Ti concentrations of M_L and M_R are slightly lower than M_E and M_A, indicating the decreasing formation temperatures.

Vanadium and Cr are multivalent and dominantly controlled by oxygen fugacity (Nadoll et al., 2014; Sun et al., 2017). Generally, V³⁺ is one of the main ions in magnetite. When fO₂ rises, V³⁺ tends to transform to less compatible V⁴⁺, which leads to the decrease of V content in magnetite (Acosta-Góngora, Gleeson, Samson, Ootes, & Corriveau, 2014). Thus, oxygen fugacity determines whether V enters the magnetite lattice, which could make V content a good record for the fO₂ during the magnetite precipitation (Toplis & Corgne, 2002). Commonly, the V content in magnetite from skarn deposit is relatively low (Nadoll et al., 2014; Shao, Chen, Ji, & Zeng, 2020). In Beiya, most samples display a V content of several tens to hundreds ppm, and M_E and M_A magnetites show a wide V concentration range (Figure 8c), which reflects that V concentration may be affected by other factor expect oxygen fugacity, such as temperature, in the hydrothermal system (Acosta-Góngora et al., 2014; Sun et al., 2017). The V contents in Beiya magnetites decrease from M_E to M_A (Figure 8c), which suggests that the fluid fO₂ increased during the main magnetite stage. Conversely, increasing oxygen fugacity may be responsible for the precipitation of large volumes of massive magnetite because increasing oxygen fugacity would improve the capacities of carrying metal by the hydrothermal fluids (Sun et al., 2017). Subsequently, the V contents increase from M_A to M_L and M_R, coupled with magnetite reduction and sulphides precipitation.

The Mg and Mn concentrations in skarn magnetite will increase due to the extensive fluid-rock reaction (Nadoll et al., 2014). Recent studies proposed that the Mg + Mn versus (Si + Al)/(Mg + Mn) diagram can reflect the degree of fluid–rock interaction (Hu et al., 2014). During the skarn formation process, the decreasing Mg + Mn contents for magnetite implies the decreasing degree of fluid–rock interaction (Hu et al., 2014). For example, from endoskarn to exoskarn in the Chenchao skarn iron deposit, magnetite from retrograde stage shows higher Mg + Mn than those from carbonate stage, revealing the decreasing degree of fluid–rock interactions (Hu et al., 2014). In Beiya, the decreasing Al, Mn, and Ga concentrations in the magnetite grains from M_E and M_A to M_L and M_R (Figure 7f and Table 1) with similar content variations indicate a decreasing degree of fluid–rock interactions.

Generally, Co content decreases in magnetite due to the occurrence of sulphides from retrograde skarn stage to carbonate stage because Co is more compatible with the sulphides (Acosta-Góngora et al., 2014). The depletion of Co of M_R compared to other three types magnetite suggests the co-precipitation of sulphides with M_R. And Ni content also can be affected by the co-precipitation minerals with magnetite (Huang et al., 2016). The Beiya magnetites have Co/Ni ratios ranging from 0.03 to 0.98, and the M_E and M_L plots tend to be more scattered. Most Co/Ni ratios of M_L (average in 0.06) and M_R (average in 0.07) are lower than M_E (average in 0.45) and M_A (average in 0.09), indicating that the forming conditions of M_L and M_R are more reduced, which may be related to the sulphide minerals deposition in the ore-forming system (Huang et al., 2016; Li et al., 2017). The Co/Ni ratios of the Beiya magnetite imply the fluid–rock reactions between ore-forming fluids and wall rocks because ore-forming fluids are considered to have a low Co/Ni ratio of ~0.01 (Bajwah, Seccombe, & Offler, 1987).

5.3 Implications of magnetite texture for ore-forming process

Studies on different types of hydrothermal deposits over the world (including skarn-type deposits, IOCG deposits, Kiruna-type iron deposits, etc.) show that the dissolution–reprecipitation process of magnetite is widespread in these hydrothermal deposits (Hu et al., 2014). Magnetite is commonly texturally and chemically altered by the dissolution–reprecipitation process. Texturally, hydrothermal magnetite that experienced dissolution–reprecipitation due to post-ore fluids replacement may show multiple generations in one magnetite (Hu et al., 2014; Hu et al., 2014; Li, Wang, et al., 2016; Shao et al., 2020). Chemically, the primary magnetite exhibits a relatively large amount of Si, Al, and Mg, while the secondary variety has a decreasing trend of composition of these trace elements, which is likely caused by trace elements leaching by later hydrothermal overprint (Hu et al., 2014). But the general variation rules of trace elements within magnetite during dissolution–reprecipitation process have not been well studied, and the behaviour of some key elements is not clear either. The trace elements in secondary magnetite could be controlled by a number of factors: (a) the composition of primary magnetite and wall rock; (b) the composition of hydrothermal fluids that replaced magnetite; and (c) physicochemical conditions, such as oxygen fugacity and temperature (Hu et al., 2014; Shao et al., 2020). Thus, trace elements in magnetite from skarn deposit could be more variable than previously thought.

Based on previous studies, the Beiya gold deposit also experienced multi-stage of magmatic-hydrothermal events (Deng et al., 2015; He et al., 2013; Sun et al., 2017). Another new magmatic event in Beiya has been identified (Li, Ding, et al., 2016; Sun et al., 2017). The M_E occurs triple-junction grain boundaries, which are a typical texture for primary magnetite (Chen, Lan, Wang, Tang, & Dai, 2018). The M_R displays a unique core–rim texture, which is possibly formed due to the dissolution–reprecipitation process (Hu et al., 2014; Shao et al., 2020). After Beiya skarn magnetite crystallized, there is at least one post-ore fluid that dissolved and modified the primary magnetite, leading to the multi-generation in one magnetite grain. Besides, the content of some trace elements in M_R and M_L is lower compared to the M_E and M_A, such as Si, Ca, Mn, and Mg, indicating the leaching of later hydrothermal fluids (Hu et al., 2014).

Mineralogical research illustrates that mineral texture is a good recorder and preserver of the physicochemical condition, and it is important to identify the preserved texture representing its formation conditions or post-formation conditions, or both (Zhu, Xie, & Li, 2019). Magnetite–haematite assemblage can reflect the redox condition of the system (Nadoll et al., 2014). In Beiya, haematite intergrown from the magnetite grain boundaries or cracks (martitization; Figure 6d) suggests that the magnetite formed in a relatively oxidized condition (Nadoll et al., 2014; Sun et al., 2017). This is consistent with the decreasing V content from M_E to M_A, indicating the increased fO₂ in the system. Thereafter, mushketovite is observed as tabular shape accompanied with pyrite and chalcopyrite, which is interpreted as a pseudomorphic replacement of haematite (Figure 6e; Zhou et al., 2017). The appearance of mushketovite symbolizes the ore-forming environment and the iron-rich hydrothermal fluids transformed from high temperature and oxidized to low-temperature and reduced features (Niu et al., 2019). In conclusion, increasing oxygen fugacity is a significant factor that leads to the deposition of large volumes of massive magnetite and partial oxidation to haematite in Beiya. Subsequent decreasing of oxygen fugacity results in haematite reduction (mushketovitization) and sulphides formation. The change of oxygen fugacity is also an important mechanism for the precipitation of gold-bearing magnetite (Zhou et al., 2017).