Comparison of Magma Oxygen Fugacity and Zircon Hf Isotopes between Xianglushan Tungsten-Bearing Granite and Late Yanshanian Granites in Jiangxi Province, South China

: Jiangxi Province (South China) is one of the world’s top tungsten (W) mineral provinces. In this paper, we present a new LA-ICP-MS zircon U-Pb age and Hf isotope data on the W ore-related Xianglushan granite in northern Jiangxi Province. The magmatic zircon grains (with high Th/U values) yielded an early Cretaceous weighted mean U-Pb age of 125 ± 1 Ma ( MSWD = 2.5, 2 σ ). Zircon ε Hf(t) values of the Xianglushan granite are higher ( − 6.9 to − 4.1, avg. − 5.4 ± 0.7) than those of the W ore-related Xihuanshan granite in southern Jiangxi Province ( − 14.9 to − 11.2, avg. − 12.5 ± 0.9), implying different sources between the W ore-forming magmas in the northern and southern Jiangxi Province. Compiling published zircon geochemical data, the oxygen fugacity ( f O 2 ) of the late Yanshanian granitic magmas in Jiangxi Province (the Xianglushan, Ehu, Dahutang, and Xihuashan plutons) were calculated by different interpolation methods. As opposed to the W ore-barren Ehu granitic magma, the low f O 2 of the Xianglushan granitic magma may have caused W enrichment and mineralization, whilst high f O 2 may have led to the coexistence of Cu and W mineralization in the Dahutang pluton. Additionally, our study suggests that the absence of late Mesozoic Cu-Mo mineralization in the Zhejiang, Jiangxi, and Anhui Provinces (Zhe-Gan-Wan region) was probably related to low f O 2 magmatism in the Cretaceous.


Introduction
South China (especially Jiangxi Province) is the main tungsten (W)-producing province in China. The early-discovered W deposits are mainly distributed in southwest Jiangxi Province, such as the renowned Xihuashan, Piaotang, and Dajishan deposits [1][2][3][4][5]. Xianglushan deposit was discovered in the 1960s in the northern Jiangxi province, a region better known for its association with world class porphyry copper (Dexing copper deposit (DCP)) and large-to medium-sized polymetallic Cu deposits [6][7][8][9][10][11][12]. The recent discovery of several super-large W deposits, such as the Zhuxi and Dahutang deposits in northern Jiangxi Province, have renewed research interest on the W metallogeny in the region and its relationship with the coeval Cu mineralization. Although the main W mineralization in both northern and southern Jiangxi Province was related to the Yanshanian (Jurassic-Cretaceous) orogeny, there are obvious metallogenic differences between them. For instance, wolframite is the main W ore mineral in southern Jiangxi Province [5,10,13], while it is mainly scheelite in northern Jiangxi Province. Moreover, in northern Jiangxi Province, W mineralization coexists locally with Cu ± Mo and the reason for such differences is still unknown.
Advances in micro-analysis have resulted in the routine in-situ measurement of key geochemical and isotopic traits of zircons grains. Over the past few decades, there has been increasing interest in the use of zircon as a mineralization pathfinder for intrusion-related mineralization [14]. In this paper, zircon data from one ore-barren (Ehu) and two fertile granites (Dahutang and Xihuashan; Table A1) are compiled, and a new zircon U-Pb ages and Hf isotope data from the Xianglushan W bearing granite in northern Jiangxi Province are presented. We compared the age, oxygen fugacity (fO2), and possible magma source of the Xianglushan granite with those of the three other granite plutons and discussed the magmatic controls on W and W-Cu mineralization in the region. Our work also provides better understanding for the W mineralization during the Yanshanian period in Jiangxi Province.

Regional Geology
The South China Block (SCB) is composed of Yangtze and Cathaysia blocks separated along the Qin-Hang belt ( Figure 1; [15,16]). Many previous studies proposed that the two blocks may have collided in the early Neoproterozoic, separated in the late Neoproterozoic, and then reassembled in the early Paleozoic (Caledonian) [3][4][5][6][7]15]. The SCB has since then experienced intensive and multiphase thermotectonic events [15,16], including those occurred in the Triassic (Indosinian) and Jurassic-Cretaceous (Yanshanian) [15,16]. Granites that formed during the Yanshanian orogeny are the most widespread, especially in the Cathaysia Block and the Qin-Hang belt. Moreover, there is a progressive coastward magmatic migration trend from the early to late Yanshanian orogeny [17][18][19].  [18,20]). Neoproterozoic and late Mesozoic (Jurassic-Cretaceous; Yanshanian) granitoids are widespread in northern Jiangxi Province, and the latter occurs mainly as stocks intruding both the former and other Precambrian rocks [21,22]. In contrast, the Nanling Range extends across the northwestern Cathaysia Block [23] and encompasses southern Jiangxi Province, southern Hunan, and western Fujian Provinces. The Nanling Range includes a Neoproterozoic schist basement and Sinian-Silurian slate [18], which are covered by Upper Devonian to Middle Triassic shallow−marine carbonate rocks, mudstones, and sandstones, and then by Upper Triassic to Paleogene terrigenous clastic rocks and volcaniclastic rocks. Two world-class W ore belts were developed in the Nanling Range [8,10,13] and northern Jiangxi Province, respectively [10]. Southern Jiangxi Province is located in the eastern Nanling Range, including Xuehuading, Xianghualing, Qianlishan, and Xihuashan deposits. The northern Jiangxi Province contains the Dahutang W-Cu, Zhuxi W-Cu, and Xianglushan W deposits ( Figure 1).
The Ehu pluton is located at about 30 km northeast of Jingdezhen and covers an area about 160 km 2 . The pluton is located on the southeastern margin of the Yangtze plate ( Figure 1). The Ehu granite intruded the low-grade meta-sedimentary rocks of the Shuangqiaoshan Group. It consists of massive medium-grained two-mica granites with an association of monzogranite-syenogranite. The rocks are mainly composed of K-feldspar (35-40%), quartz (30%), plagioclase (24%), biotite (5%), and muscovite. Most of the plagioclase grains are sericitized and biotite is partially replaced by chlorite. Accessory minerals include mainly zircon, apatite, epidote, and Fe-Ti oxides. Moreover, the Ehu granites are devoid of Cu (Au)-Mo or Sn-W mineralization [26].

Zircon Morphology and Texture
Zircon separation was conducted on a ~2 kg crushed rock sample (XLS01-1) at the Hongxing Geological Laboratory (Langfang, China). After heavy liquid and electromagnetic separation, zircon grains with better crystal shape and transparency were picked under the microscope. The internal structure of zircon grain was observed via cathodoluminescence (CL) imaging and transmitted-/reflected-light microscopy. All of the CL imaging were conducted at the Wuhan Sample Solution Analytical Technology Co. Ltd. (Wuhan, China). Zircon CL images were obtained using an Analytical Scanning Electron Microscope (JSM-IT100) connected to a GATAN MINICL system. The imaging condition was 10.0-13.0 kV accelerating voltage of electric field and 80-85 µA current of tungsten filament.

Zircon U-Pb Dating
LA-ICP-MS zircon U-Pb dating and trace element analysis were simultaneously conducted at the same laboratory as zircon CL imaging. The analyses were performed with a GeolasPro laser ablation system that consists of a COMPexPro 102 ArF excimer laser (193 nm wavelength and maximum 200 mJ energy) and a MicroLas optical system. Helium was used as a carrier gas, and argon as the make-up gas that mixed with helium via a T-connector before entering the ICP. A "wire" signal smoothing device was included in this laser ablation system [29,30]. The laser spot size and frequency were set to 32 µm and 5 Hz, respectively, and Plešovice zircon was used as the external standard. The obtained Plešovice (338.6 ± 1.1 Ma) ages are consistent with the value reported by [31]. The off-line selection and background-analyzed signal integration, trace element calibration, and time-drift correction were performed with the in-house (CUG, Wuhan, China) ICPMSDataCal software (Version 10.9) [30]. Common Pb correction was carried out using with the measured 204 Pb contents [32]. Concordia diagrams and weighted mean calculations were plotted with the Isoplot/Ex_ver3 [33].

Zircon Hf Isotopic Analyses
In-situ Hf isotope analysis was conducted using a Neptune Plus MC-ICP-MS (Thermo Fisher Scientific, Dreieich, Germany) coupled with a Geolas HD excimer ArF laser ablation system (Coherent, Göttingen, Germany) at the same laboratory as the zircon U-Pb dating. Analysis conditions include 44 µm spot size, 8 Hz laser repetition, and 5.3 J/cm 2 energy density, and other details are as described in Hu et al. [30]. Spot locations of the Hf isotopic analyses are shown in Figure  3. The analysis requires careful correction of isobaric interferences on 176 Hf (e.g., 176 Yb and 176 Lu). It is observed that the mass fractionation of Yb (βYb) is not constant over time, and the βYb obtained from the solution introduction is unsuitable for the measurements [34]. The βYb miscalculation would affect the 176 Hf/ 177 Hf results. In this study, we used the βYb values directly obtained (real-time) from the zircon grains. Additionally, the 173 Yb/ 171 Yb and 179 Hf/ 177 Hf values were applied to estimate the mass bias of Yb (βYb) and Hf (βHf), which were normalized to 173 Yb/ 171 Yb (1.13268) and 179 Hf/ 177 Hf (0.73255) [35] with an exponential correction. Meanwhile, interference of 176 Yb on 176 Hf was corrected by measuring the interference-free 173 Yb and utilizing 176 Yb/ 173 Yb (0.79639) [35] to calculate 176 Yb/ 177 Hf. Similarly, the relatively minor interference of 176 Lu on 176 Hf was corrected by measuring the interference-free 175 Lu intensity and used 176 Lu/ 175 Lu (0.02656) to estimate 176 Lu/ 177 Hf. Since Yb and Lu have similar elemental behaviors, βYb was applied to calculate the mass fractionation of Lu. The offline processing of analytical data (e.g., mass bias calibration, sample selection, and blank signal) were performed with the ICPMSDataCal software [30]. Our analyses yielded weighted mean 176 Hf/ 177 Hf ratios of 0.2820172 ± 0.0000060 for the GJ-1 zircon standard and 0.2823080 ± 0.0000035 for the 91500zircon standard.

Zircon Log fO2 Ratios, Ce and Eu Anomaly Estimation
Zircon has high closure temperature and is resistant to weathering or hydrothermal alteration. In general, Ce in magma has two valence states (Ce 4+ and Ce 3+ ). Compared to Ce 3+ , Ce 4+ has the same ionic radius and valence state as Zr 4+ in the zircon lattice. Therefore, Ce 4+ (instead of Ce 3+ ) is compatible in magmatic zircon, which thus show strong positive Ce anomaly in chondritenormalized REE (Rare Earth element) pattern. Various zircon Ce-based oxygen fugacity (fO2) indicators were developed to assess the magmatic redox conditions. In particular, Trail et al. [36] proposed an equation (Equation (1)) that can directly calculate the absolute value of oxygen fugacity. This equation has been widely adopted in many studies on the genesis of world-class porphyry Cu deposits: where (Ce/Ce*)D is the zircon Ce anomaly estimated from the partition coefficients and T is the absolute zircon crystallization temperature. Recent studies suggested [37,38] that substitution of Ti, Si 4+ , and Zr 4+ in zircon lattice depends primarily on temperature. As a result, the titanium content can estimate zircon crystallization temperature, if the TiO2 and SiO2 activities in the melt at the time of crystallization are well constrained. Therefore, Equation (2) proposed by Ferry and Watson [37] was used to calculate the magma temperature at the time of zircon crystallization.
Where α SiO and α TiO represent the Si and Ti activity, respectively. Notably, the studies of natural samples and experiments by Trail et al. [36] suggested that Ce anomalies in the magma can be calculated by the following approximation: CHUR is the abbreviation of chondrite uniform reservoir, where (Ce/Ce*) CHUR represents the Ce anomalies normalized to the chondritic uniform reservoir (CHUR; [39,40]). However, the measurement of Ce/Ce* = CeN/(LaN*PrN) 1/2 (subscript N indicates chondrite normalization) anomaly is difficult because La and Pr are very difficult to be measured precisely. Moreover, the two elements are susceptible to contamination by tiny melt and titanite inclusions that are common in zircon [38]. This has led some authors to use a ratio between Ce and a more abundant REE as a proxy for Ce enrichment or depletion (such as Ce/Nd; e.g., [40][41][42]. Loard et al. [43] argued that Ce/Ce* can be estimated based on Equation (5), in which Sm and Nd were less affected by the inclusions and can be measured more precisely. Thus, determination of the Ce anomaly and fO2 for the magmas using the Sm-Nd interpolation can yield more robust fO2 values.
Furthermore, zircon EuN/EuN* ratios can also evaluate the magmatic oxygen fugacity, because Eu 2+ cannot substitute into zircon due to its cationic size and charge [42]. However, the redox effect on zircon EuN/EuN* ratios is complicated by the strong partitioning of Eu 2+ into other minerals, notably plagioclase [36,38,40]. Plagioclase crystallization can deplete the melt in Eu relative to Sm and Gd [43]. Hence, zircon Eu anomalies are not only influenced by redox, but also by the plagioclase abundance. In view of this, Eu anomaly is not used to assess the fO2 of granitic magmas in this study.

Zircon U-Pb Age
Zircon grains separated from XLS01-1 sample are colorless, euhedral transparent, and about 50 to 150 µm long. Most zircon grains have fine oscillatory zoning and some are sector-zoned ( Figure 3). A total of 39 zircon grains from the sample were analyzed (Table 1) and their U and Th contents and Th/U ratios are of 255 to 8210 ppm, 244 to 1268 ppm, and 0.15 to 1.12, respectively (Table 1), resembling typical magmatic zircons (Th/U > 0.1, [12]). U-Pb ages of the zircon grains are highly consistent (122 to 129 Ma), which yielded a weighted-mean age of 125 ± 1 Ma (n = 40, MSWD = 2.5; Figure 4), representing the crystallization age of the biotite granite XLS01-1.

Temperature-Redox Conditions
Calculated Ti-in-zircon temperatures of the W-related granites from the three deposits and the Ehu ore-barren granite are listed in Table A1. The activities of TiO2 and SiO2 were estimated to be 0.7 and 1, respectively. The Ti-in-zircon temperatures are 666 to 786 °C (avg. 699 ± 32 °C) for the Xianglushan biotite granite, 709 to 848 °C (avg. 745 ± 35 °C) for the Dahutang granite, 654 to 890 °C (avg. 727 ± 51 °C) for the Xihuashan granite, and 654 to 890 °C (avg. 727 ± 33 °C) for the Ehu granite.
The log fO2 values for the W-related and barren granites were listed in Table A1 and illustrated in Figures 6 and 7. It is noted that the log fO2 values were calculated by the La-Pr and Nd-Sm interpolation methods, respectively. As shown in Figure 6a, most data from the Xianglushan granite are located below those of the Ehu ore-barren granite. Log fO2 values range from FMQ −12.01 to 4.76 (avg. −7.97 ± 3.73) for the bearing granite, and from FMQ −10.13 to +7.59 (avg. −2.36 ± 4.32) for the barren one (Figure 6b). Although the results are broadly consistent with the previous redox estimates for these zircons [37], our results indicate a slightly oxidizing environment. It is also noted that the log fO2 range of Ehu granite in Figure 6c is narrower than that in Figure 6a. The results estimated by Equation (3) range from FMQ −13.00 to +3.61 (avg. −7.16 ± 4.5) for the ore-bearing granite, and from FMQ −6.01 to +5.54 (avg. −1.93 ± 2.46) for the barren granite (Table A1; Figure 6d).  (c) logfO2 vs. temp diagram, where logfO2 value was calculated with the method of Trail et al. [34], and CeN* with the Sm-Nd fitting method; (d) Histogram of oxygen fugacity. Data are listed in Table   A1.
As shown in Figure 7, although the Dahutang granite zircon data are partially overlapped with those from the Xihuanshan and Xianglushan granites, zircon grains from the Dahutang ore-bearing granite still have the highest logfO2 values regardless of the calculation method. The La-Pr interpolation approach yielded logfO2 values of FMQ −9.34 to +4.76 (avg. −0.37 ± 3.63) for the Dahutang granite, FMQ −11.70 to +3.33 (avg. −2.80 ± 4.33) for the Xihuashan granite, and FMQ −12.01 to 4.76 (avg. −7.97 ± 3.73) for the Xianglushan granite (Figure 7a,b). In contrast, the Sm-Nd interpolation method obtained a narrower logfO2 range (Figure 7c,d), and higher values (albeit some overlapping) for the Dahuatang (FMQ −7.33 to +5.90; avg. −0.79 ± 3.27) and Xihuashan (FMQ −10.29 to +4.97; avg. −1.92 ± 3.67) granites than the Xianglushan granite (FMQ −13.00 to +3.61; avg. −7.16 ± 4.54) (Figure 7; Table A1). Notably, unlike the La-Pr interpolation, the Sm-Nd fitting method does not require accurate measurement of La or Pr and is thus considered to be more robust. Meanwhile, some studies [25] further suggested that zircon REEs have a concave-downward (rather than linear) chondrite-normalized pattern. Hence, neither of the two methods can accurately determine Ce* and would result in under-/over-estimation of the true Ce*.

Timing of Magmatism and its Related W Mineralization
LA-ICP-MS zircon U-Pb dating suggests that the Xianglushan granite was formed at 125 ± 1 Ma, coeval to the W mineralization (scheelite Sm-Nd age: 121 ± 11 Ma [23]). Moreover, the molybdenite Re-Os age and 40 Ar- 39 Ar dating of muscovite are also consistent with the U-Pb zircon age of 125 ± 1 Ma for the biotite granite. On the basis of the LA-ICP-MS zircon U-Pb, molybdenite Re-Os, and muscovite 40 Ar-39 Ar ages [9,23], it is concluded that W mineralization at the Xianglushan deposit is genetically associated with the biotite granite. Magmatism and mineralization in the Xianglushan deposit occurred during the Early Cretaceous. Additionally, published age data indicate two late Mesozoic (Yanshanian) magmatic event in Jiangxi Province (Table 3; Figure 8): The first Late Jurassic-Early Cretaceous events occurred mainly at 160-140 Ma, e.g., the emplacement of ore-related granite (porphyries) at the Zhuxi W-Cu (148-152 Ma) and Xihuashan W deposits (158-161 Ma) [13].The second event occurred mainly at 135-120 Ma, e.g., the Xianglushan granite (ca. 125 Ma; this study). It is noteworthy that both magmatic events were reported at the Dahutang W deposit (Table 3), and both events may have been ore-related [49,50]. In summary, the Yanshanian magmatism and its related W mineralization in southern Jiangxi Province occurred in a relatively confined period of time, while the magmatism and related W-Cu-(Mo) mineralization in northern Jiangxi Province are characterized by occurring as multi-phases at a wider age span.  Table 3.

ɛHf(t) Variation and Ore Material Source
Previous studies demonstrated that the W mineralization is closely related with Cu-(Mo) mineralization in northern Jiangxi Province, which is uncommon worldwide outside of South China [2,23,47]. Tungsten is an a lithophile element due to a valence of +6 in nature. Currently, over 20 Wbearing minerals (notably wolframite and scheelite) have been identified in nature [2,10,47,55,56]. The increase of oxygen content between the core and the mantle leads to the separation of tungsten from the core and its entry into the mantle [47,57]. Meanwhile, tungsten is an incompatible element and tends to accumulate in the crust during the crust mantle process and evolution. O'Neill et al. [57] reported that partition coefficients for W between silicate and Fe-rich metal will be highly increased under reduced condition. Unlike W, the oxidized magma is beneficial for Cu-(Au)-Mo mineralization through controlling the valence of sulfur. Thus, W and Cu tend to be enriched in the crust and mantle, respectively [10,12,18].
In this study, W-bearing granites from both the northern and southern Jiangxi Province have negative εHf(t) values (−14.9 to −2.4). Moreover, the Hf two-stage model age vary from 1085 to 1634 Ma and the Hf two-stage model of the individual granite is relatively uniform in age, indicating that these granites may have mainly crustal source [47]. Meanwhile, the ɛHf(t) values of ore-related granites in northern Jiangxi Province are clearly higher than those in southern Jiangxi Province. As shown in Figure 5c, compared to the ore-related granites in southern Jiangxi Province, those in northern Jiangxi Province have the younger TDM2 age which indicates the major source difference between the northern and southern magmas. This may have caused by partial melting of different metamorphic substrates [12,24,56]. According to the whole-rock geochemical data, granites in southern Jiangxi Province have higher SiO2, but lower Al2O3, TiO2 and MgO, and significantly lower P2O5 contents. The A/CNK-A/NK diagram suggests that W ore-related granites from both southern and northern Jiangxi Province are peraluminous. In addition, the ore-bearing granites in southern Jiangxi Province have higher Rb/Sr, but lower Zr/Hf, LREE/HREE and Eu/Eu* than their northern Jiangxi counterparts. By comparing the granite whole-rock 87 Sr/ 86 Sr and ɛNd(t) values from northern Jiangxi Province and those of the Neoproterozoic Shuangqiaoshan Group, Su and Jiang [48] proposed that the former may have partly originated from the latter, which contains much higher contents of W (avg. 9.13 ppm) and Cu (avg. 38.1 ppm) than the average continental crust (W: 1 ppm, Cu: 27 ppm; [47]). Comparatively, both the ore-related granites and wall-rock sequences in southern Jiangxi Province have high W background contents [47,48], which may have contributed some ore-forming materials for the (super)-large W mineralization in the region. Notably, differences in Sr/Sr, Nd and Hf isotopes of granites in north and south do indicate probable differences in the source [8,47]. But the geochemical characteristics as lower Zr/Hf, Eu/Eu, Al2O3, TiO2, and MgO or higher SiO2 and Rb/Sr might be most probably related to differences in the fractionation of these magmas [13,43,58], that is, this probably indicates that granites from the south are more evolved that those from the north.

Oxygen Fugacity Variation and Implications
Pirajno [59] proposed that the fO2 dependency of mineralization increases in the Sn-W-Mo-Cu-Mo-Cu-Au sequence, while the Fe dependency increases in the Mo-Sn-W-Cu-Mo-Cu-Au sequence. Nevertheless, the relationship between fO2 and W mineralization is still not fully understood [18]. Some workers believed that low fO2 is beneficial to W mineralization, whereas many others suggested that fO2 plays little role in the W mineralization. Although W mineralization shows little dependence on magma fO2 in view of geochemical affinity [60], large-scale W mineralization is always closely associated with reduced granites [18,47]. In our study, no matter which calculation method is used, the fO2 of the Xianglushan granite is always lower than that of the ore-barren Ehu granite. This suggests that lower fO2 may have been beneficial for W mineralization in northern Jiangxi, e.g., at Xianglushan. It may be explained by that low fO2 facilitates W enrichment in silicate melts during source melting and magmatic differentiation. However, as shown in Figure 7, Dahutang granite has the highest fO2, and could be interpreted as having different magma sources. Based on the Hf isotope evidence, the Dahutang ore-bearing granite in northern Jiangxi Province was probably sourced from arc-type materials (of the Shuangqiaoshan Group), which commonly have high logfO2 values (NNO +1 to +3). Copper mineralization is generally associated with oxidized magmas [60][61][62][63] and can explain the coexistence of Cu and W mineralization in the northern Jiangxi Province.
Magma oxygen fugacity has been widely accepted as the most important control of Cu-Mo-Au mineralization (e.g., [5,8,47,64]). Many recent studies showed that high fO2 granitic magma is the key for Cu mineralization in northeastern Jiangxi Province. For example, the ore-related granites in the Dexing and Tongcun PCDs have likely high fO2 [11]. Meanwhile, Qiu et al. [26,62] suggested that the low magma fO2 found in several Mo ore-related and ore-barren porphyries in western Zhejiang Province may have contributed to the Cretaceous Cu-Mo mineralization gap in the Zhe-Gan-Wan region. This hypothesis is supported by our study, as shown in Figure 9 and Table 4, Nd-Sm interpolation approach yielded logfO2 values of FMQ −10.29 to +11.87 (avg. +3.54 ± 4.02) for the Late Jurassic granites (145-170 Ma) FMQ −13.00 to −5.55 (avg. −3.69 ± 4.70) for the Early Cretaceous granites (120-145 Ma). It shows that fO2 gradually decreased from Jurassic to Cretaceous (Figure 9).

Tectonic Implications
Although the origin of strong oxidation has been still argued, a broad consensus has been reached that high oxidization is associated with subduction zone [18,36,40]. Sun et al. [18,59] proposed that subduction zone can release fluids to elevate oxygen fugacity. The closer distances from subduction zone are, the more fluids contribute and the higher oxygen fugacities are. For example, many hydrous (3−5 wt % water) arc magmas have high fO2, ranging from NNO + 1 to NNO + 3 [18,60]. This phenomenon likely because of high amounts of dehydration-released fluid containing a lot of oxidized materials (i.e., Fe 3+ , Mn 4+ , S 6+ , and C 4+ ) in subduction zones [18,60]. Additionally, the tectonic evolution of South China in late Mesozoic remains controversial for a long time [7,25,59]. A variety of tectonic models have been presented to address the Late Mesozoic largescale magmatism and mineralization in South China, with most models invoking subduction of the paleo-Pacific plate [2,4,62]. Based on the drifting direction of the Pacific plate before 125 Ma [18,59] and the age distribution of magmatic rocks and mineralization zonation in the Late Jurassic to the Early Cretaceous in southern China like that in the South America. Sun et al. [59] proposed that the paleo-Pacific plate was subducting from NE to SW during ca. 180-125 Ma, and from SE to NW after 125 Ma in South China. This model may explain the differences in oxygen fugacity between the Late Jurassic and the Early Cretaceous granites in South China. According to this model, the Late Jurassic granites (such as Dexing granodiorite porphyry) are input more oxidized materials from the subduction than that of Early Cretaceous (such as Xianglushan biotite granite). Hence, it is observed that fO2 of granites gradually decreased from Jurassic to Cretaceous (Figure 9). Furthermore, Dahutang granite is closer from subduction zone than that of Xihuashan granite and Dahutang granite has the higher oxygen fugacity than that of Xihuashan granite.

Implications for Zircon as an Indicator
Zircons are widely distributed in igneous rocks and have stable geochemical properties. They faithfully record the information of zircon crystallization (i.e., fO2). Moreover, with the development of analytical technology, researchers have obtained a lot of REE (rare earth element) as well as zircon age data. Ce and Eu are variable valence elements, whose valence state are affected by the redox conditions of magma. Unlike CeN/CeN*, EuN/EuN* in zircon is generally affected by the crystallization of plagioclase [18,40]. Hence, most researchers use Ce anomalies of zircon/melt distribution coefficients (Di) to estimate the oxygen fugacity. A series of oxygen fugacity barometers have been developed based on Ce anomalies to indicate the redox conditions of magma since 2002 [37,43,64]. However, these methods produced a wide range of Ce/Ce* (fO2) values which vary by up to 3 orders of magnitude for a single rock in some studies [43]. In this study, all zircon CeN/CeN * ratios from selected granites also show a wide range (Figures 6 and 7). The cause of this phenomenon has not been well understood [40,43]. One possibility is that some tiny inclusions (i.e., monazite, apatite, and titanite) frequently can be detected in zircon. Meanwhile, as shown in Figure 6, no matter which calculation method is used, the fO2 of the Xianglushan granite is always lower than that of the orebarren Ehu granite. Therefore, it is suggested that zircon fO2 may be still used as an indicator to discriminate ore-bearing and barren granites in areas of W mineralization.

Conclusions
1. LA-ICP-MS zircon U-Pb dating of the Xianglushan biotite granite yielded an Early Cretaceous age (125 ± 1 Ma). Age compilation indicates that the magmatism and W-Cu-(Mo) mineralization in northern Jiangxi Province are characterized by being multiphase, while the magmatism and W mineralization in southern Jiangxi Province occurred mainly in the Middle to Late Jurassic (165-150 Ma).
2. The ore-related granites in northern Jiangxi Province have a younger TDM2 age and clearly higher ɛHf(t) values than those in southern Jiangxi Province, which seem to indicate a major source difference between the northern and southern granitic magmas in Jiangxi Province.
3. Compared with the coeval ore-barren Ehu granites, the low fO2 of the ore-related Xianglushan granite may have caused the W enrichment and mineralization, whilst the high fO2 of the Dahutang granite may have facilitated the coexistence of Cu and W mineralization. Zircon fO2 may be still used as an indicator to discriminate ore-bearing and barren granites in some cases.
4. Variation of oxygen fugacity among different granites may support a model that the Paleo-Pacific plate was subducting southwestwardly, as proposed in some previous studies.

Acknowledgments:
We are grateful to the L.-L. Yuan for assistance with the sampling, and H. Song for the fruitful discussion. We appreciate Wei Gao for helping with the LA-ICP-MS analyses.

Conflicts of Interest:
The authors declare no conflict of interest.
Appendix A Table A1. Crystallization temperatures, Ce anomalies, and magma oxygen fugacity of zircon grains from the Xianglushan, Dahutang, and Xihuashan granites.