Introduction

Porphyry Cu–Au deposits are one of the most important ore deposit types worldwide and provide 75% of the Cu and 20% of global Au production (Sillitoe 2010). Porphyry Cu–Au deposits are mainly found at the margins of converging plates, such as island and continental arcs (Richards 2003, 2015). The primary magmas of subduction-related porphyry Cu–Au deposits are mainly mafic magmas formed by partial melting of the mantle wedge induced by dehydration of the subducting plate (Cooke et al. 2005; Sillitoe 2010; Richards 2015; Chen and Wu 2020). However, only a few subduction-related porphyries form giant Cu–Au deposits. The formation of porphyry Cu–Au deposits is affected by many factors, including: (1) the enrichment of metallic elements in the deep primary magma (Bell et al. 2009); (2) the magma oxidation state (Richards 2003); (3) the timing of sulfide saturation in the magma (Park et al. 2015, 2019; Hao et al. 2017); (4) the magma water concentration (Halter et al. 2005); (5) the ascending pathways of the ore-bearing magmas (Richards 2003); (6) the abundance of S and Cl in the magma (Grondahl and Zajacz 2022); and (7) the formation and migration of ore-bearing hydrothermal fluids (Mungall 2002; Wilkinson 2013). Copper and Au are highly compatible in sulfide melts and are thus strongly affected by sulfide saturation. The solubility of S in magmas is sensitive to the magma oxidation state (Jugo 2009). This study investigated how the magma oxidation state and sulfide saturation history of arc magmas control the metal fertility of arc magmas and mineralization potential.

The Gangdese belt has experienced a complex and protracted geological evolution, involving northward subduction in the Mesozoic Neo-Tethys Ocean and the Cenozoic collision between India and Eurasia (Yin and Harrison 2000; Zhang et al. 2021). This evolution has generated a 1500-km-long porphyry mineralization belt in the Tibetan Plateau with Cu resources of >56 million tons (Zheng et al. 2021). Neo-Tethyan oceanic subduction resulted in two periods of magmatism, which produced a large amount of Cretaceous and lesser amounts of Jurassic arc-related igneous rocks. The Jurassic rocks are associated with the giant Xiongcun porphyry Cu–Au deposit (~240 Mt with 0.4 wt% Cu; 173–161 Ma; Tang et al. 2010), whereas no large porphyry deposits associated with Cretaceous arc magmatism have been identified. The reason for this remains unknown.

Numerous Cenozoic porphyry Cu–Au deposits formed in a continental post-collisional setting have been discovered in the Gangdese belt (Hou et al. 2009, 2015; Yang et al. 2014, 2015; Richards 2015). Based on studies of Miocene continental post-collisional porphyry Cu–Au deposits, it has been concluded that these are different from typical arc porphyry deposits in terms of the deep geodynamic processes and magma sources (e.g., Hou et al. 2009, 2015). The magma and metal sources for porphyry Cu–Au deposits in a post-collisional setting are thought to result from the remelting of juvenile mafic lower crust that contains sulfides (Hou et al. 2015; Wang et al. 2019). However, recent studies have suggested that the contribution from the lower crust has been overestimated (Zhang et al. 2022a). It has also been proposed that the magmas associated with post-collisional porphyry Cu–Au deposits were formed by fractionation crystallization of lithospheric mantle-derived K-rich magmas (Chang and Audétat 2022).

Late Cretaceous mafic–ultramafic igneous rocks containing variable amounts of sulfides are found in the Milin area of the Gangdese belt and are thought to be the typical juvenile lower crust formed during subduction in the Neo-Tethys Ocean. We investigated the zircon chemistry and whole-rock platinum-group element (PGE), Au, and S contents of the Milin rocks. We used these data to constrain the magmatic oxygen fugacity (fO2) state and S saturation history during partial melting of the mantle and magmatic evolution, and how this affected mineralization. This provides new insights into the potential of the Late Cretaceous Gangdese arc magmatism to form porphyry Cu–Au deposits and whether lower crustal remelting can provide sufficient metallic elements for the formation of porphyry Cu–Au deposits.

Geological setting and sample descriptions

The Tibetan Plateau in southwest China is the largest and highest plateau in the world (Fig. 1a). It is divided into five blocks, including the Kunlun, Songpan–Ganzi, Qiangtang, Lhasa, and Himalayan blocks, from north to south (Fig. 1b). These blocks are separated by the Kunlun, Jinsha, Bangong Lake–Nujiang, and Yarlung–Zangbo suture zones, respectively, which record the closure of the Paleo-Asian and Paleo-, Meso-, and Neo-Tethys oceans, respectively (Yin and Harrison 2000; Zhang et al. 2021).

Fig. 1
figure 1

(a) Tectonic map of China (modified after Yuan et al. 2010). (b) Tectonic framework of the Tibetan Plateau. (c) Distribution of Mesozoic and Cenozoic igneous rocks in the Lhasa terrane (after Zhang et al. 2021). BNSZ = Bangong–Nujiang suture zone; YZSZ = Yarlung–Zangbo suture zone

Full size image

The Lhasa Block is a microcontinental block that separated from the Gondwana supercontinent and comprises mainly Precambrian basement, Paleozoic–Mesozoic sedimentary, and Mesozoic–Cenozoic igneous rocks (Zhu et al. 2011, 2013). It has experienced the continental collision between India and Eurasia in the early Cenozoic (65–55 Ma; e.g., Mo et al. 2003; Zhu et al. 2015), collision between the Lhasa and Qiangtang blocks in the Late Jurassic to Early Cretaceous (e.g., Pan et al. 2006; Zhu et al. 2013), and subduction in the Bangong Lake–Nujiang Tethys Ocean (to the south) and Yarlung–Zangbo Tethys Ocean (to the north) during the Mesozoic (Zhu et al. 2013). The Lhasa Block is a vast tectonomagmatic belt that has experienced multiple stages of magmatism. The Lhasa Block extends for 2500 km from east to west and is divided into northern, central, and southern sub-terranes by the Shiquan River–NamTso Mélange Zone (SNMZ) and the Luobadui–Milashan Fault (LMF) (Fig. 1b; Zhu et al. 2013). The northern Lhasa Block consists mainly of juvenile crust (Hou et al. 2015), Triassic–Cretaceous strata (sandstone, limestone, slate, and chert), and abundant volcanic rocks (Pan et al. 2006). The central Lhasa Block comprises Neoproterozoic metamorphic crystalline basement, which is covered by widely distributed Carboniferous–Permian clastic metasedimentary rocks (Hu et al. 2005). In addition, there are large areas of Jurassic–Cretaceous volcanic–sedimentary rocks, and a small amount of Ordovician, Silurian, and Devonian strata (Zhu et al. 2011, 2013). The southern Lhasa Block comprises sparse Precambrian crystalline basement, Upper Triassic–Lower Jurassic strata of the Yeba Formation (clastic sedimentary and volcanic rocks), and the Upper Jurassic–Cretaceous Sangri Formation (sandstone, slate, mudstone, and limestone) (Zhu et al. 2013).

The Gangdese belt is located at the southern margin of the central and southern Lhasa blocks and extends for 1500 km from east to west, with a width of 100–200 km from north to south (Fig. 1c). It records northward subduction of the Neo-Tethyan oceanic lithosphere and the collision between India and Eurasia (e.g., Yin and Harrison 2000; Ding and Lai 2003, Ding et al. 2014; Chu et al. 2006; Zhang et al. 2021). As a result, Jurassic volcanic rocks and abundant Cretaceous igneous rocks were generated during subduction, and large-scale magmatism (including K-rich magmatism) occurred during the collisional and post-collisional stages. There is a giant porphyry ore deposit associated with the Jurassic igneous rocks, but no large porphyry deposits associated with the Cretaceous magmatism have been identified. Unlike typical porphyry Cu–Au deposits worldwide, most porphyry deposits formed during the post-collisional stage in the Gangdese belt.

The eastern Gangdese arc consists mainly of three formations (i.e., the Bala, Gongga, and Rouqiecun formations) and four batholiths and complexes (i.e., the Linzhi, Kanniang, Wolong, and Lilong batholiths/complexes) (Fig. 2). The Bala Formation consists of bistagite, amphibolite, and biotite amphibole schist. The Gongga and Rouqiecun formations consist mainly of low-grade (greenschist-facies) metamorphic rocks represented by quartz and two-mica schists. The Linzhi Complex consists of Neoproterozoic to early Cenozoic garnet-bearing gneiss, amphibolite, and quartzite (Dong et al. 2011; Guo et al. 2011). The Kanniang Complex consists mainly of garnet-bearing gabbro. The Wolong batholith comprises Late Cretaceous granodiorite and granite (Sun 2021). The Lilong batholith is an ultramafic–intermediate intrusion and consists of hornblendite, gabbro, diorite, and granodiorite. These rocks have typical cumulate textures and are thought to be the typical juvenile lower crustal cumulates (Zhang et al. 2010; Sun 2021). The studied samples were collected from the Lilong batholith in western Milin County (Figs. 1c and 2).

Fig. 2
figure 2

Geological map of the Milin area, eastern Gangdese arc, showing sample locations

Full size image

The studied samples are mainly olivine–pyroxene hornblendite, pyroxene hornblendite, hornblendite, and hornblende gabbro (Fig. 3a–c). The rocks all have cumulate textures and contain disseminated sulfides such as chalcopyrite, pentlandite, pyrrhotite, and pyrite. Sulfides occur mainly at the intervals of silicate mineral grains. Rounded sulfide droplets of pyrrhotite, chalcopyrite, pentlandite, and pyrite are widely distributed in hornblende in the various lithologies (Fig. 3d–h). This indicates the sulfides are magmatic in origin. The olivine–pyroxene hornblendite consists of olivine (25–30 vol.%), pyroxene (10–15 vol.%), and hornblende (50–55 vol.%) (Fig. 3a). The pyroxene hornblendite consists of pyroxene (25–30 vol.%) and hornblende (65–70 vol.%). The pyroxenes have two main forms, with one being subhedral and the other being inclusions in hornblende. Hornblendite is the main sulfide-rich cumulate rock and consists mainly of hornblende (~95 vol.%) (Fig. 3b). The hornblende gabbro consists mainly of pyroxene (30–35 vol.%), plagioclase (45–50 vol.%), and hornblende (~15 vol.%) with minor biotite (Fig. 3c).

Fig. 3
figure 3

Representative photomicrographs of the Milin mafic–ultramafic cumulates showing (a) olivine–pyroxene hornblendite, (b) hornblendite, (c) hornblende gabbro, and (d–h) disseminated sulfide in the silicate minerals. Ol = olivine; Hbl = hornblende; Cpx = clinopyroxene; Pl = plagioclase; Ccp = chalcopyrite; Pn = pentlandite; Po = pyrrhotite; Py = pyrite

Full size image

Analytical methods

Whole-rock major elements were analyzed using X-ray fluorescence (XRF) spectrometry at the ALS Laboratory Group, Guangzhou, China. The powdered samples (0.7 g) and fluxing agent (7 g; Li2B4O7) were fused and cooled into glass discs for analysis. The precision is <±5%. Trace elements were measured using a Perkin–Elmer ELAN-DRC-e inductively coupled plasma mass spectrometer (ICP–MS) at the State Key Laboratory of Ore Deposit Geochemistry (SKLODG), Institute of Geochemistry, Chinese Academy of Sciences, Guiyang, China. The powdered samples (50 mg) were dissolved in HF + HNO3 in high-pressure Teflon bombs for 8 h at ~190°C. The solutions were then evaporated to dryness and dissolved in 2% HNO3 for ICP–MS analysis. The international standard AGV-2 was used to monitor the data accuracy, and the precision is <±10% for all elements.

Two hornblende gabbro samples from the Milin area were selected for zircon separation at the Langfang Regional Geological Survey, Hebei Province, China. We observed the morphology of the zircons in cathodoluminescence (CL) images and in transmitted and reflected light and selected zircon grains for U–Pb dating and trace element analysis. This process was conducted by laser ablation ICP–MS (GeoLas Pro 193 nm ArF excimer laser and Agilent 7500x ICP–MS) at the SKLODG. The laser beam diameter was 32 μm, with an 8 Hz repetition rate and energy density of 10 J/cm2. Helium was used as the carrier gas. Data for each zircon were obtained for 60 s after 20 s of background data acquisition. The zircon standard 91500 was used as an external standard for the age corrections. The standard materials NIST 610, BHVO-2G, BCR-2G, and BIR-1G were used to correct the zircon trace element data. The zircon U–Pb isotope ratios and trace element contents were obtained using ICPMSDataCal 10.8 software (Liu et al. 2009). Concordia plots were obtained using Isoplot 3.0 (Ludwig 2003). The common Pb correction followed the method of Andersen (2002).

In situ zircon Lu–Hf isotopic compositions were obtained using a Neptune Plus III multi-collector ICP–MS coupled to a New Wave 213 nm laser ablation system at SKLODG. Analyses employed a laser spot size of 44 μm and He as a carrier gas, with a repetition rate of 10 Hz and energy density of ~5.3 J/cm2. During the analyses, the Penglai standard zircon yielded a weighted-mean 176Hf/177Hf ratio of 0.282904 ± 0.000012, which is in good agreement with the recommended Hf isotope ratio for this standard (176Hf/177Hf = 0.282906 ± 0.000010; Li et al. 2010).

Gold and S were analyzed at the ALS Laboratory Group, Guangzhou, China. The samples (25–30 g) were weighed and then digested in aqua regia in a high-density polyethylene bottle. Gold was analyzed using an Agilent 7900 ICP–MS. The detection limit for Au was 0.1 ppb. The samples were placed in an induction furnace at 1350°C, and S released in the form of SO2 was introduced into an infrared detection system. The S contents were measured according to the change in the energy received by the C–S analyzer (CS844 LECO). The detection limit for S was 0.001 wt%.

The PGEs were measured using an improved digestion technique and isotope dilution (ID)–ICP–MS at SKLODG (Qi et al. 2011). The powdered samples (~5 g) were added to ~5 mL of water in a polytetrafluoroethylene (PTFE) beaker. Subsequently, 25 mL of HF was added and the beakers were placed on a hotplate to remove silicates. After the solutions had evaporated, 5 mL of HF and 15 mL of HNO3 were added and the samples were placed in an oven for 48 h at 190°C. The samples were then evaporated to dryness on a hotplate, and 5 mL of HCl was added to remove the residual HF + HNO3 in the sample. After drying, the residues were dissolved in 40 mL of HCl. The solutions were centrifuged at 2800 rev/min for 5 min. The upper part of the solutions was used to collect the PGEs by Te co-precipitation. The precipitates were dissolved in aqua regia and diluted to 10 mL, then the PGEs were extracted on a mixed ion exchange column. Finally, the filtered solutions were evaporated down to ~3 ml for ICP–MS analysis. Platinum, Pd, Ru, and Ir were measured using isotope dilution, and the Rh contents were calculated using 194Pt as an internal standard. The international standard WGB-1 and TDB were used to monitor the data accuracy. The detection limits are 0.004 ppb for Ir, 0.008 ppb for Ru, 0.006 ppb for Rh, 0.014 for Pt, and 0.012 for Pd.

Results

Zircon U–Pb age, Hf isotope, and trace element data

The in-situ zircon U–Pb and Hf isotope and trace element data are listed in Tables S1–3. The zircon grains from the two hornblende gabbro samples exhibit weak oscillatory zoning and have crystal lengths of 50–200 μm, with length/width ratios of 1 to 2. Twenty-nine zircon grains from sample 20ML-26 yielded a weighted-mean 206Pb/238U age of 84.3 ± 1.4 Ma (MSWD = 1.11; Fig. 4a). These zircons have initial 176Hf/177Hf ratios of 0.282959–0.283164, with corresponding εHf(t) values of 8.37–15.81 (Fig. 5). Eighteen zircon analyses of sample 20ML-32 yielded a weighted-mean 206Pb/238U age of 86.7 ± 1.9 Ma (MSWD = 6.4; Fig. 4b), which represents the crystallization age. The zircons have initial 176Hf/177Hf ratios of 0.283058–0.283149, with corresponding εHf(t) values of 11.94–15.18 (Fig. 5).

Fig. 4
figure 4

(a, b) Zircon U–Pb concordia diagram and (c, d) zircon chondrite-normalized REE patterns for the Milin hornblende gabbro

Full size image
Fig. 5
figure 5

Plots of εHf(t) versus age for zircons from the Milin samples and Jurassic igneous rocks from the Gangdese arc, southern Tibet. The εHf(t) data for the Jurassic rocks are from Ji et al. (2009), Chu et al. (2011), Hou et al. (2015), and Xie (2019)

Full size image

The remaining zircons have high heavy rare earth element (REE) and low light REE contents, with high total REE contents (ƩREE = 47–1004 ppm), and marked negative to small positive Eu anomalies (EuN/EuN* = 0.12–1.38) (Fig. 4c–d). The REE patterns and high Th/U ratios (0.13–0.67) suggest these are magmatic zircons (Hoskin and Schaltegger 2003). Previous studies have shown that zircon Ce4+/Ce3+ ratios are closely related to the magmatic fO2 (Ballard et al. 2002; Loucks et al. 2020). The zircon Ce4+/Ce3+ ratios are calculated from the method of Ballard et al. (2002) based on Ce contents in zircon and melt. Some mineral inclusions (e.g., rutile and apatite) occur in the zircon grains, and the inclusions can strongly affect the zircon trace element contents. Therefore, some anomalous results (e.g., La > 0.1 ppm and Ti > 20 ppm) were excluded in this study (Zhu et al. 2022). The zircons from the 20ML-26 and 20ML-32 have low Ce4+/Ce3+ ratios of 21–90 and 33–76, with low estimated mean fO2 values of ΔFMQ–3.5±0.5, and ΔFMQ–1.4±0.2 based on the method of Loucks et al. (2020), and these samples have a total mean fO2 values of ΔFMQ–1.8±0.5.

Major and trace element geochemistry

The whole-rock geochemical compositions of the mafic–ultramafic cumulates are listed in Table S4. The loss-on-ignition (LOI) values of the samples are generally <3 wt% (1.22–3.00 wt%). The samples have low SiO2 contents of 41–50 wt%. The variations between major elements and MgO are shown in Fig. 6. SiO2 and Na2O + K2O increase initially with decreasing MgO until 17 and 15 wt%, respectively, and then decrease (Fig. 6a and b). Fe2O3T decreases initially with decreasing MgO until 17 wt% and then increases (Fig. 6c). Al2O3 increases with decreasing MgO (Fig. 6d). Ni and Co decrease with decreasing MgO (Fig. 6e and f). The major element variations reflect the accumulation of silicate minerals (e.g., olivine, pyroxene, and hornblende) during early magmatic evolution.

Fig. 6
figure 6

Plots of (a) SiO2, (b) Na2O+K2O, (c) Fe2O3T, (d) Al2O3, (e) Ni, and (f) Co versus MgO for the Milin mafic–ultramafic cumulates

Full size image

On primitive mantle-normalized trace element (Fig. 7a) and chondrite-normalized REE diagrams (Fig. 7b), all the samples have similar patterns, with enrichments in large-ion lithophile elements (LILEs; e.g., Rb, Ba, and K) and depletions in high-field-strength elements (HFSEs; e.g., Nb, Ta, Zr, and Hf). The REE patterns are flat ([La/Yb]N = 0.86–2.79), and the samples have no obvious Eu anomalies (Eu/Eu* = 0.8–1.07; mean = 0.96).

Fig. 7
figure 7

Chondrite-normalized REE patterns and primitive mantle-normalized trace element diagrams for the Milin mafic–ultramafic cumulates. Normalization values were taken from Sun and McDonough (1989)

Full size image

Chalcophile element and sulfur contents

The PGE, Cu, Au, and S contents of the analyzed rocks are listed in Table 1. The Milin mafic–ultramafic cumulates contain 0.16–22.43 ppb Pd (mean = 5.7 ppb), 0.13–15.53 ppb Pt (mean = 3.94 ppb), 0.01–0.42 ppb Rh (mean = 0.12 ppb), 0.01–0.16 ppb Ru (mean = 0.06 ppb), and 0.005–0.33 ppb Ir (mean = 0.1 ppb). The total PGE contents are highly variable (0.34–29.91 ppb). The Pt and Pd contents increase initially with decreasing MgO until ~17 wt% and then decrease (Fig. S1a–b). The Ir and Ru contents decrease with decreasing MgO (Fig. S1c–d). The Cu and Au contents of the samples are 12.6–432 ppm (mean = 151 ppm) and 0.6–10.2 ppb (mean = 3.62 ppb), respectively. The Cu contents exhibit no clear trend with MgO (Fig. S1e). The Au contents increase initially with decreasing MgO until ~17 wt% and then decrease (Fig. S1f). The Cu/Pd ratios of the samples are 5.43 × 103 to 1.12 × 106 and are higher than those of the mantle (103 to 104; Barnes and Maier 1999). These ratios are negatively correlated with Pd (Fig. 8b). The S contents vary from 150 to 7800 ppm (mean = 1360 ppm). The correlation between PGEs, Cu, and Au with S is weak (Fig. S2), which is a common phenomenon in sulfide-bearing rocks (Maier et al. 2004; Lightfoot et al. 2012; Keays and Tegner 2015). This weak correlation can be attributed to the different partition coefficients and R-factors (mass ratio of silicate to sulfide liquid; Campbell and Naldrett 1979) between sulfide melt and silicate magma during sulfide saturation.

Table 1 PGE (ppb), Au (ppb), Cu (ppm) and S (ppm) contents of the Milin mafic–ultramafic cumulates
Full size table
Fig. 8
figure 8

(a) Primitive mantle-normalized PGE patterns and (b) plot of Pd versus Cu/Pd for the Milin mafic–ultramafic cumulates. Primitive mantle values are from Barnes and Maier (1999)

Full size image

Discussion

Reduced arc magmas during the Cretaceous

The mantle wedge oxygen fugacity exerts a key control on arc magma evolution and mineralization in subduction zones. Owing to metasomatism by subduction-derived melts or fluids, mantle wedges are generally considered to be more oxidized than the sources of depleted mantle-derived magmas, such as mid-ocean ridge basalts (i.e., MORBs; Ballhaus 1993; Brounce et al. 2014; Richards 2015). For example, Ballhaus (1993) showed that the fO2 of a mantle wedge ranges from FMQ+1 to FMQ+3 using the olivine–orthopyroxene–spinel equilibrium approach. Brounce et al. (2014) showed that the fO2 of island arc primary magmas varied between FMQ+1 and FMQ+1.6 using the Fe3+/ΣFe ratio of olivine-hosted melt inclusions.

Recent studies have shown that the fO2 was high in arc magmas generated in the early subduction stage (e.g., Jurassic) of the Neo-Tethys Ocean, which is consistent with typical arc magmas worldwide. For example, Chen et al. (2019) reported Ce4+/Ce3+ ratios of 126–3444 for zircons from the Bima Formation arc volcanic rocks (195–165 Ma), and Zou et al. (2015) and Xie et al. (2018) reported that zircons from the Xiongcun porphyry deposit (179–164 Ma) have Ce4+/Ce3+ ratios of 197–3737 (Table S5). These high zircon Ce4+/Ce3+ ratios are indicative of a high fO2 in these arc magmas (Ballard et al. 2002; Trail and Bruce Watson 2012). We estimated the fO2 (Fig. 9) of these Jurassic arc magmas using published zircon data (Wei 2014; Zou et al. 2015; Ma et al. 2017; Xie et al. 2018; Chen et al. 2019), based on the method of Loucks et al. (2020). The Jurassic Bima and Yeba formation volcanic rocks yielded a mean fO2 of ΔFMQ+1.3 and ΔFMQ+1.0, respectively. The Xiongcun porphyry deposit yielded a mean fO2 of ΔFMQ+1.3. These results are consistent with the high fO2 of the Jurassic arc magmas. In contrast, zircons from the Milin mafic–ultramafic cumulates have low Ce4+/Ce3+ ratios (21–90), implying the Late Cretaceous arc magmas had a low fO2. The estimated fO2 of the Milin cumulates ranges from ΔFMQ–3.5 to ΔFMQ–1.0, with a mean of ΔFMQ–1.8±0.5 (Fig. 9), which is much lower than for the Jurassic arc magmas. Therefore, the fO2 of the arc magmas in the Gangdese belt decreased significantly from the Jurassic to Late Cretaceous.

Fig. 9
figure 9

Redox state of Jurassic–Late Cretaceous magmatism in the Gangdese arc. Data are from Zou et al. (2015), Ma et al. (2017), Xie et al. (2018), and Chen et al. (2019)

Full size image

The decrease in the fO2 may have been due to the mantle source or magmatic evolution. During magmatic evolution, variations in fO2 are caused by fractional crystallization and crustal contamination. Carmichael (1991) proposed that during fractional crystallization, the fO2 variations depend on the Fe3+/ΣFe ratio of the residual magma. If the separated minerals cause the Fe3+/ΣFe ratio of the residual magma to increase, then the residual magma becomes more oxidized. However, most studies have shown that fractional crystallization does not lead to significant changes in fO2 (de Hoog et al. 2004; Crabtree and Lange 2012; Grocke et al. 2016). For example, when MgO decreases from 10 to 7, and to 5 wt% in MORBs, the Fe3+/ΣFe ratio only increases by 0.015 and 0.03, respectively (Cottrell and Kelley 2011; Shorttle et al. 2015). Previous studies have shown that garnet and hornblende crystallized during fractionation of the parental magmas to the Milin mafic–ultramafic cumulates (Sun 2021; Zhang et al. 2021). The effects of garnet and hornblende fractionation on magma fO2 are controversial (King et al. 2000; Tang et al. 2018; Lee and Tang 2020; Zhang et al. 2022b; Holycross and Cottrell 2023). Some studies have suggested that garnet and hornblende crystallization will increase the Fe3+/ΣFe ratio and fO2 in the residual magma (Tang et al. 2018; Lee and Tang 2020; Zhang et al. 2022b). Consequently, the low fO2 of the Milin cumulates cannot have been caused by garnet and hornblende fractionation. The crystallization of magnetite will reduce the fO2 of magma. The Xiongcun porphyry deposit and Jurassic volcanic rocks (e.g., Bima and Yeba formation) in the Gangdese experience magnetite differentiation (Zou et al. 2015; Xie et al. 2018; Chen et al. 2019), but their fO2 are still higher than those of the Milin samples. This means that the differentiation of magnetite cannot be the cause of the low fO2 in Milin area. Therefore, the low fO2 of the Late Cretaceous arc magmas was not caused by fractional crystallization.

The addition of oxidizing or reducing crustal material can obviously change the fO2 of magma. For example, the magma of the Duke complex is considered to be reduced by the addition of C-bearing country rocks (Thakurta et al. 2008). However, the Milin mafic–ultramafic cumulates have obvious depleted Hf isotopic compositions (8.37–15.81), similar to those (10.2–17.7; Fig. 5) of the Jurassic arc magmas, indicating that crustal contamination was negligible because crustal material usually has enriched Hf isotope compositions. As such, crustal contamination did not cause the decrease in fO2 from the Jurassic to Late Cretaceous. In fact, the low fO2 may be inherited the mantle source. For instance, Hu et al. (2023) reported the Troodos mantle ophiolite have reduced fO2 values. Furthermore, the chalcophile element depletion in the primary magmas indicates that the primary magmas had a low fO2, implying that the decrease in fO2 may have been inherited from the mantle source. Recent studies have shown that several global oceanic anoxic events (OAEs) occurred worldwide during 120–86 Ma (i.e., in the Cretaceous; Huber et al. 2002; Chen et al. 2011). During OAEs, organic matter-rich black shales were widely deposited in marine and terrestrial environments. Therefore, the low fO2 of the Late Cretaceous mantle source may be caused by the addition of reducing fluid by organic matter-rich sediments, but this still needs further study.

First stage of chalcophile element depletion—residual sulfide in the mantle during partial melting

The fO2 of a magmatic system has a prominent role on S solubility (Jugo 2009), which in turn controls the behavior of chalcophile and siderophile elements (Lee et al. 2012). The S content of the depleted mantle is generally <200 ppm (Nielsen et al. 2014; Ding and Dasgupta 2017). However, owing to metasomatism during subduction, S-containing fluid or melt will enter the mantle wedge and increase the S content of the mantle wedge (300–500 ppm; de Hoog et al. 2001; Wallace and Edmonds 2011; Richards 2015; Li et al. 2022). As a variable-valence element, the valence state (S2− and S6+) of S is markedly affected by the fO2. Sulfide (S2−) and sulfate (S6+) are mainly present under low-fO2 (e.g., ≤FMQ) and high-fO2 (e.g., >FMQ+2; Jugo 2009) conditions. The solubility of S can increase by 10 times with increasing fO2 (Jugo 2009).

Typical arc magmas are produced by 8%–20% partial melting of the mantle wedge (Kelley et al. 2006). At a high fO2, only a small percentage of partial melting is needed to dissolve all the sulfide in the mantle. The chalcophile elements, such as PGEs, Au, and Cu, which are hosted in the sulfide, will thus enter the basaltic magma during partial melting. This process results in the formation of a primary arc magma with metal fertility. Within these primary arc magmas, Copper, Pd and Au can reach up to 200–300 ppm, 20–30 ppb and 8 ppb, respectively (Sun et al. 2004; Dale et al. 2012; Park et al. 2013; Chiaradia 2014; Richards 2015). These concentrations are expected to have Cu/Pd ratios similar to those found in primitive mantle (103 to 104). For example, Pt and Pd in Tongan primary arc magmas (MgO > 9 wt%) are up to 21.0 and 21.2 ppb, respectively, and the Cu/Pd ratios are 4.6 × 103 to 1 × 104 (Dale et al. 2012; Park et al. 2013). Sulfide saturated from this magma would be enriched in PGEs and have low Cu/Pd ratios. For example, the sulfide saturated from magmas of the Tolbachik volcano in the Kamchatka arc contain up to 299 ppm Pd and 115 ppm Pt and have low Cu/Pd ratios (9 × 102 to 4.3 × 104; Zelenski et al. 2017).

The studied samples are mafic–ultramafic cumulates and were early-formed cumulates generated from arc magmas. These rocks contain a variable amount of sulfides (Fig. 3) but are characterized by PGE depletions and high Cu/Pd ratios (5.43 × 103 to 1.12 × 106; Fig. 8). Due to the higher sulfide/silicate partition coefficients for PGEs than Cu, such as PGEs (103 to 105; Bezmen et al. 1994; Zhang and Li 2021; 105 to 106; Mungall and Brenan 2014) and Cu (102 to 103; Crocket et al. 1997), the Cu/Pd ratios of sulfide saturated from a magma would be lower than the initial magma ratios. Therefore, the primary magma of the Milin lower crust would have had higher Cu/Pd ratios than the sulfide-bearing mafic–ultramafic rocks and much higher values than those of the mantle. The high Cu/Pd ratios of Milin sulfide-bearing cumulates imply that the primary magmas were characterized by depletions in PGEs and other chalcophile elements (e.g., Cu and Au). This, together with the low total PGE contents, are inconsistent with sulfide saturated from a chalcophile element-enriched magma generated by partial melting of mantle with a high fO2. In fact, the fO2 (ΔFMQ–1.8±0.5) of the Milin primary magma is more reduced than typical arc magmas. Under such low-fO2 conditions, a high degree of partial melting is required to dissolve all the sulfide in the mantle. At a reasonable degree of partial melting, S-saturated arc magma would be generated and residual sulfide would remain in the mantle. Therefore, the arc magma would have low PGEs contents and high Cu/Pd ratios. This is the case for the Milin cumulates. The estimated chalcophile element contents of the primary arc magma generated by partial melting of the mantle wedge under reduced conditions are shown in Fig. 10. The results show that partial melting of 15% would yield a primary magma with 63 ppm Cu, 0.29 ppb Au, 0.029 ppb Pd, and 0.036 ppb Pt at FMQ–1 (Table S6). Therefore, the Milin primary magmas are characterized by moderate Cu depletion and large Au and PGE depletions.

Fig. 10
figure 10

Mantle melting models. The high temperature during partial melting of the mantle leads to a decrease in the Cu and Au partition coefficients, and thus we assumed that the partition coefficients for Cu, Au, Pd, and Pt between sulfide liquid and silicate melt are 8 × 102, 7 × 103, 2.85 × 105, and 4 × 105, respectively (Mungall and Brenan 2014; Li and Audétat 2015). The initial S, Cu, Au, Pd, and Pt contents in the primitive mantle were 300 ppm, 30 ppm, 1 ppb, 4 ppb, and 7 ppb, respectively (Barnes and Maier 1999; Zhao et al. 2022). The S solubility was 1400 ppm under ΔFMQ–1 condition (Zhao et al. 2022)

Full size image

The variation of Cu and Pd in these rocks, as shown in Fig. 11, can be modeled using a two-stage sulfide saturation approach with different R-factors. We assume that the arc magmas contained 10 ppb Pd, 5 ppb Au, and 100 ppm Cu if all the Cu-rich Fe sulfide liquid in the mantle had been dissolved in the magma. This assumption is based on the average contents of chalcophile elements in chalcophile elements-undepleted arc magma worldwide (Sun et al. 2007; Dale et al. 2012; Park et al. 2013). The partition coefficients between sulfide liquid and silicate for Au, Cu, and Pd are 8 × 103, 1 × 103, and 2.85 × 105, respectively (Mungall and Brenan 2014; Li and Audétat 2015). The modeling results indicate that during the first stage sulfide saturation, which corresponds to the sulfide retained in the mantle during partial melting, the primary magmas become depleted in chalcophile elements, containing only 0.07 ppb Pd, 1.0 ppb Au, and 66.7 ppm Cu under an R-factor of 2000. The variation of Pd and Cu in the Milin cumulates was formed during the second stage sulfide saturation from the chalcophile elements-depleted primary magma, with R-factors ranging roughly from 103 to 105 (Fig. 11).

Fig. 11
figure 11

Model calculation of the variation of Pd with Cu in the Milin mafic–ultramafic cumulates. Cyan solid points present the sulfide liquids segregated from PGE-rich basaltic magma with 10 ppb Pd and 100 ppm Cu under variable R-factor values. Red solid points show the sulfide liquids segregated from the PGE-depleted magma with 0.07 ppb Pd, and 66.7 ppm Cu under variable R-factor values. The sulfide/silicate partition coefficients for Cu and Pd are 1 × 103 and 2.85 × 105, respectively (Mungall and Brenan 2014; Li and Audétat 2015). Percentages of sulfide in the rocks are labeled as 0.1%, 1% and 10%

Full size image

In summary, the low fO2 in the mantle source resulted in the presence of residual sulfides in the mantle during partial melting, and thus the Milin primary magmas were depleted in chalcophile elements.

Second stage of chalcophile element depletion caused by early sulfur saturation

The S solubility in the Milin primary magma was low due to the low fO2, and thus the primary magma was saturated in sulfide before it left the mantle. In general, the S solubility increases as pressure decreases (Wendlandt 1982; Mavrogenes and O’Neill 1999). Therefore, as pressure decreases during magma ascent into the lower crust, the magma will become sulfide-undersaturated. As shown in Fig. 3, the early crystallized minerals contain sulfide droplets, suggesting the magma reached sulfide saturation in the early stages of magmatic differentiation. The S solubility is mainly affected by pressure, temperature, magma composition, and oxygen fugacity. The fO2 of the magma was too low to decrease the S solubility. The magma Fe content has an important effect on the S solubility, and thus magnetite crystallization typically leads to sulfide saturation. However, the Milin cumulates are the products of early crystallization of magma, and magnetite crystallization was negligible.

For a typical primary arc basaltic magma, the S solubility is about 1400 ppm at 2 GPa under reduced condition and, when the pressure drops to 1 GPa, the S solubility increases to 1700 ppm (Fortin et al. 2015). Owing to S being an incompatible element in olivine, clinopyroxene, and other early liquidus minerals, S contents will increase with fractional crystallization in the residual magma. The Milin primary magma was S-saturated, and thus the S content in the primary magma was equal to its solubility (i.e., the Milin primary magma contained 1400 ppm S). Only 20% fractional crystallization is required to increase the S content in the magma to 1700 ppm, which then results in S saturation. In addition, the S solubility decreases by about 400 ppm for every decrease of 100°C at 1 GPa (Holzheid and Grove 2002), which suggests that a temperature decrease could also have resulted in sulfide saturation. Therefore, we propose that as fractional crystallization proceeded and temperature decreased during early magmatic evolution, S became saturated in the Milin basaltic magmas. As such, metallic elements were sequestered by the saturated sulfide, which further reduced the metal contents of the residual magma. We estimate that the residual basaltic magma after S saturation in the early stage of magmatic differentiation contained 33-66 ppm Cu, 0.13–0.93 ppb Au, and 0–0.02 ppb Pd. These contents are much lower than those of typical arc basaltic magmas (Cu = 93–135 ppm; Au = 3.84–6.06 ppb; Pd = 7.61–13.7 ppb; Pt = 2.80–7.61 ppb; Sun et al. 2007; Park et al. 2013).

Mineralization implications for the Gangdese belt

Late Cretaceous arc magmatism

The Gangdese belt has experienced multiple stages of magmatism and porphyry Cu–Au mineralization. The large Xiongcun porphyry Cu–Au deposit was associated with Jurassic arc magmatism (Tang et al. 2010). There are abundant Late Cretaceous arc igneous rocks in the Gangdese belt, but no large porphyry deposits associated with these have been found. There are many factors involved in the formation of porphyry Cu–Au deposits, but it is important that there are enough metallic elements in the magma. Our results show that residual sulfide in the mantle during partial melting was due to the low fO2, which resulted in the primary magma being moderated depleted in Cu and highly depleted in Au and PGEs. Furthermore, the primary magma reached sulfide saturation in the lower crust during the early stages of magmatic differentiation. Abundant chalcophile elements were sequestered in the saturated sulfide in the lower crust. The residual magma after S saturation contained 33–66 ppm Cu, 0.13–0.93 ppb Au and 0–0.02 ppb Pd, much lower than typical arc basaltic magmas.

It has been found that some large Cu porphyries are formed in association with magma with lower content of Cu. Recent studies have shown that the redissolution of sulfide can form magmatic–hydrothermal fluids containing large amounts of metallic elements and thus provide sufficient materials for the formation of porphyry Cu–Au deposits (Wilkinson 2013; Mungall et al. 2015; Bai et al. 2020). However, the exsolution of hydrothermal fluids from magma is physical release and it more likely to occurs in the shallow crust. Although saturated sulfides can be transported by ascending andesitic arc magmas (Heinrich and Connolly 2022), the magma from the lower crust must pass through multi-level magma chambers to reach the shallow crust (Fig. 12; Annen et al. 2006), which will reduce the magma flow rate and lead to sulfide being precipitated in the multi-level magma chambers. As a result, the Milin lower crustal sulfides were difficult to transfer to the shallow crust and could not be dissolved in aqueous fluids.

Fig. 12
figure 12

Schematic diagram of the evolution of the Milin arc magmas in the Gangdese arc. Under the low fO2 conditions, partial melting of the mantle wedge with metasomatism by subduction-related fluid or melt produced a sulfide-saturated primary arc magma, with depletions in chalcophile elements due to the residual sulfide in the mantle. The primary arc magmas reached sulfide saturation owing to fractional crystallization proceeded and the temperature decreased during the early stages of magmatic differentiation. In this case, the chalcophile elements were sequestered into the early formed sulfide, which further reduced the chalcophile element contents in the residual basaltic magma. Both these factors hindered the formation of large Late Cretaceous porphyry Cu–Au deposits in the Gangdese belt

Full size image

Based on our results, we propose that sulfide saturation in the early stages of magmatic differentiation and residual sulfide in the mantle during partial melting hindered the formation of large porphyry Cu–Au deposits associated with the Late Cretaceous arc magmatism in the Gangdese belt (Fig. 12).

Implications for the metal sources of the post-collisional porphyry Cu–Au deposits

Because of the early saturation and accumulation of sulfide during magmatic differentiation, the Milin lower crustal cumulates contain 1361 ppm S and 151 ppm Cu; i.e., much higher than the contents of their mantle source. In the Gangdese belt, numerous Cenozoic porphyry Cu deposits formed in a post-collisional setting. It has long been proposed that the ore-bearing magmas formed by remelting of thickened, sulfide-rich, juvenile lower crust (Richards 2009; Hou et al. 2015; Wang et al. 2019). However, recent studies have shown that the contribution from the lower crust to the porphyry Cu deposits was overestimated (Zhang et al. 2022a). To assess the importance of lower crust in the formation process of the post-collisional porphyry Cu deposit, we conducted remelting simulations of the Milin lower crustal mafic cumulates. Remelting of mafic–ultramafic cumulates generally produces andesitic magmas. We assumed the temperature and pressure were 800°C–1000°C and 1 GPa, respectively. Under these conditions, sulfide exists mainly as monosulfide solid solution (MSS; Li and Audétat 2015). Given the partition coefficients of Cu between silicate minerals (olivine, pyroxene, and hornblende) and melt range of 0.05 to 0.2 (Fellows and Canil 2012; Liu et al. 2014), the bulk D was set to 0.1 for silicate assemblages. The partition coefficients of Cu between MSS and melt are estimated according to Li and Audétat (2015). The S solubility of the andesitic magma are calculated from the equations of Jugo et al. (2010), Fortin et al. (2015), and Zajacz and Tsay (2019). More detailed parameters of the model are listed in the caption of Fig. 13. According to the results of partial melting experiments (e.g., Beard and Lofgren 1991; Qian and Hermann 2013; Gao et al. 2016), a water-rich mafic crustal material can reach a maximum of 30%–50% partial melting at 800 °C. The Milin mafic–ultramafic rocks are rich in hornblendes and represents a water-rich lower crust. Moreover, Lu et al. (2015) reported that additional water was added during the partial melting process of the Gangdese Miocene lower crust. Therefore, a high degree (50%) of melting is assumed in our model. The estimated contents of metals in the magma generated by remelting of the Milin lower crustal cumulates are listed in Table S7 and shown in Fig. 13.

Fig. 13
figure 13

Remelting models for Cu in the Milin mafic–ultramafic cumulate. The average contents of the Milin mafic–ultramafic cumulate are 1361 ppm S and 151 ppm Cu. We assumed the temperature and pressure were 800°C–1000°C and 1 GPa. Under these conditions, the sulfide exists mainly as monosulfide solid solution (MSS; Li and Audétat 2015; Sun 2021). DCu = 2100, 1900, 1700 and 830, 740, 660 at 1 GPa, 800°C, 1000°C and FMQ, FMQ+1, FMQ+2 conditions, respectively. The S solubility of the andesitic magma are 230, 410, 600 and 560, 1000, 2400 ppm under 800°C, 1000°C and FMQ, FMQ+1, FMQ+2 conditions, respectively (Jugo et al. 2010; Fortin et al. 2015; Zajacz and Tsay 2019)

Full size image

The results show that the Cu content in the melt produced by 800°C and <50% partial melting of the lower crust is 19–26 ppm (Fig. 13a and b). Because of the lower magma S solubility compared to the S content in the lower crust, the melt generated is always S-saturated at 800°C. Thus, the Cu content in the melt do not increase markedly with an increasing degree of partial melting, even under high fO2 condition (Fig. 13a and b). On the contrary, there are higher Cu (48–112 ppm) content in the initial melt under 1000°C condition, which are higher than those (Cu ~40 ppm) of typical andesitic arc magmas (~3 wt% MgO; Sun et al. 2007; Park et al. 2015; Richards 2015). This is due to higher S solubility of magma and low partition coefficient of Cu between MSS and magma at 1000°C (Holzheid and Grove 2002; Li and Audétat 2015). The increased S solubility enables the melt to dissolve more sulfide. This, together with the low partition coefficient, results in a higher metal content in the melt compared to the content generated at low temperatures.

Our modeling show that the remelting of the lower crust can generate a magma with fertile metals under conditions of high temperature and high fO2, and magma with barren metals under low temperatures even with high fO2. Indeed, there are Miocene ultrapotassic magmas with high fO2 in the Gangdese belt (Li et al. 2020), and their underplating and addition may contribute to the high fO2 of the magma generated by the lower crustal remelting (Hou et al. 2004; Wang et al. 2014). Therefore, the high temperature is crucial to whether the lower crust remelting can provide abundant metal to match the Gangdese Miocene post-collisional porphyry Cu deposit. Previous studies have shown that the formation temperature of ore-bearing porphyry in the post-collision stage of the Gangdese belt is relatively low. For instance, Lu et al. (2015) reported that zircon saturation temperatures of the ore-bearing porphyries ranging from 680 to 780°C. Wu (2016) calculated that the crystallization temperature of hornblende in different porphyry deposits was between 700 and 800°C. However, it is unclear whether the lower crust source was remelted at high temperatures, which is depend on the heat budgets allow by the ultrapotassic magmas. If the temperature of the remelting is not high enough, additional metal may be contributed to the formation of the Gangdese Miocene post-collisional porphyry Cu deposits. Consequently, further investigation on temperature for source melting is probably needed.

Conclusions

The studied mafic–ultramafic cumulates (86.7–84.3 Ma) in the lower crust were the products of Late Cretaceous subduction in the Neo-Tethys Ocean. The cumulates have low Ce4+/Ce3+ ratios (21–90) and low fO2 values (mean ΔFMQ–1.8±0.5) that are lower than those of the Jurassic arc magmatism (126–3737; ΔFMQ+1.0 to ΔFMQ+1.3), suggesting the fO2 of the arc magmas in the Gangdese belt decreased from the Jurassic to Late Cretaceous. The reduced nature of the magmas cannot be explained by fractional crystallization or crustal contamination, suggesting inheritance from a mantle source.

Under low-fO2 conditions, residual sulfide remained in the mantle during partial melting, resulting in significant depletions in chalcophile elements in the Milin primary arc magmas. In the lower crust, sulfide saturation in the magma during the early stages of magmatic differentiation led to a further decrease in chalcophile element content in the residual magmas. Both these factors hindered the formation of large Late Cretaceous porphyry Cu–Au deposits in the Gangdese belt. Remelting of the Milin sulfide-rich cumulates can produce a Cu-rich andesitic magma only under high temperature and high-fO2 conditions, and a magma with low Cu content under low temperature even high fO2 conditions. Therefore, the temperature of the lower crust remelting is a crucial factor in determining whether it can supply sufficient metals to match the Gangdese Miocene post-collisional porphyry Cu deposit, and further investigations are necessary.