Gold precipitation in hydrothermal systems is traditionally attributed to supersaturation of gold due to decreasing gold complex stability triggered by changes in physicochemical conditions of the ore fluid. However, ultrahigh-grade gold veins in orogenic (shear zone related) gold deposits can contain kilograms per tonne of gold or more, in marked contrast to the typically very low gold concentrations (tens of parts per billion) in fluid. The gold mineral assemblage is commonly restricted to native gold and/or Au-(Ag)-tellurides and occurs in micro-fractures of sheared quartz veins. Textural and compositional characterization of such assemblages, coupled with hydrothermal diamond anvil cell experiments and heating-freezing experiments, provides evidence for an alternative ultrahigh-grade gold enrichment mechanism via growth of polymetallic melt droplets induced by quartz fracturing. We propose that polymetallic melt droplets of Au-Ag-Te-Bi–rich composition form through adsorption-reduction of metal complexes on fractured quartz surfaces, where surface silanol groups and hydrogen serve as reductants. The melt droplets subsequently grow by catalyzing reduction of metal complexes and absorbing metals from fluids percolating in the fractured quartz network. The mobile and reactive polymetallic melt droplets can repeatedly react with the fluid on protracted quartz fracturing and efficiently continue to scavenge gold from multiple pulses of gold-undersaturated ore fluids.

The precipitation of gold in hydrothermal ore deposits is generally attributed to the supersaturation of gold in aqueous or aqueous-carbonic fluids. However, it appears doubtful whether ore fluids can ever reach bulk saturation with respect to gold because of the relatively low concentration of dissolved gold species compared to gold solubilities (e.g., Simmons and Brown, 2006; Pokrovski et al., 2014; Guo et al., 2018). This suggests there may be precipitation mechanisms that are not controlled by bulk gold solubilities, especially for the formation of ultrahigh-grade gold veins. This is because gold supersaturation triggered by processes such as boiling and fluid mixing would be accompanied by the precipitation of quartz, calcite, and other common vein minerals, which strongly dilute gold grade. This contrasts with the observation that gold minerals, e.g., native gold or Au-(Ag)-tellurides, are frequently a major or dominant component in micro-fractures of ultrahigh-grade gold quartz veins that can contain kilograms or more of gold per tonne.

Alternatively, gold far below bulk saturation can be fixed at mineral surfaces through adsorption-reduction reactions (e.g., Bancroft and Hyland, 1990; Widler and Seward, 2002). Highly efficient adsorption-reduction of gold complexes on fluid-mineral interfaces could be potentially achieved through the formation and growth of polymetallic melt droplets, which can catalyze the decomposition of metal complexes and subsequently absorb metals from solution into the polymetallic melt droplets (e.g., Trentler et al., 1995; Daeneke et al., 2018). Polymetallic melt droplets, dominated by low-melting-point chalcophile elements (i.e., Zn, Ga, As, Se, Ag, Cd, In, Sn, Sb, Te, Hg, Tl, Pb, and Bi; Frost et al., 2002) and Au, have been reported in hydrothermal gold deposits of different types (e.g., Ciobanu et al., 2006; Cook et al., 2009; Cockerton and Tomkins, 2012; Hastie et al., 2020; Jian et al., 2021, 2022). Moreover, the precipitation of polymetallic melt droplets from hydrothermal fluids through adsorption-reduction mechanisms on pyrrhotite has been experimentally proven (Tooth et al., 2011).

None of these previous studies has, however, addressed the role of quartz in polymetallic droplet formation and gold enrichment, even though quartz commonly is the dominant mineral in gold ores. Furthermore, the ore fluids from which polymetallic melt droplets form have never been systematically investigated due to the scarcity of coexisting fluid inclusions suitable for characterization and analysis. Accordingly, empirical evidence for the exact temperatures and other physicochemical conditions of polymetallic droplet formation is largely lacking.

Using a combination of scanning electron microscope (SEM) imaging and energy-dispersive X-ray spectrometry (EDS), heating-freezing experiments, and hydrothermal diamond anvil cell experiments, we present evidence for entrapment of polymetallic melt droplets in quartz, constrain the conditions of melt formation, and propose a model of ultrahigh-grade gold enrichment through polymetallic droplet growth induced by quartz fracturing and generation of hydrogen and silanol groups during earthquake faulting.

The investigated ultrahigh-grade gold ore samples (>1000 g/t Au) were collected from the S16 gold-bearing quartz vein (34°24′N, 110°35′E), Xiaoqinling gold district, central China (Jian et al., 2015, 2021). The polymetallic melt droplets studied appear as polycrystalline inclusions in quartz. They coexist with low-salinity H2O-CO2 fluid inclusions (5.3–9.7 wt% NaCl equivalent, Th (total homogenization temperature) 293–400 °C; Tables S1–S2 in the Supplemental Material1) along healed fractures in quartz and in some cases were trapped within fluid inclusions (Fig. 1). The polymetallic inclusions invariably consist of multiple mineral phases, including petzite (AuAg3Te2), calaverite (AuTe2), native gold, tellurobismuthite (Bi2Te3), rucklidgeite (PbBi2Te4), altaite (PbTe), and chalcopyrite, with minor galena, bornite, hessite (Ag2Te), buckhornite (AuPb2BiTe2S3), and volynskite (AgBiTe2) (Fig. 2). Two hundred and seventy-five individual polymetallic inclusions from 10 assemblages were investigated by SEM-EDS, indicating that they are dominated by Au, Ag, Te, and Bi, with minor Pb, Cu, Fe, and S. The 10 inclusion assemblages give the following average bulk chemical composition: 31.0 ± 3.1 wt% Au, 18.3 ± 2.8 wt% Ag, 40.2 ± 2.7 wt% Te, 6.2 ± 2.1 wt% Bi, 3.6 ± 2.6 wt% Pb, 0.5 ± 0.3 wt% Cu, 0.1 ± 0.2 wt% Fe, and 0.2 ± 0.2 wt% S (Table S3). Backscattered electron images, photomicrographs, schematic drawings, and calculated compositions of the 275 polymetallic inclusions from the 10 assemblages can be found in Figure S1 and Table S4 (see footnote 1).

On heating by a Linkam THMSG600 heating-freezing microscope stage, the polymetallic inclusions, consisting mainly of Au, Ag, Te, and Bi, started melting at phase boundaries or triple point junctions at temperatures as low as 180 °C (Fig. 3; Fig. S2), far lower than the Au-Ag-Te eutectic temperature (at 304 °C; Cabri, 1965) and the Bi-Te-Au eutectic temperature (235 °C; Prince et al., 1990). The low initial melting temperature of minerals within the inclusions is due to the co-presence of multiple elements (i.e., Au, Ag, Te, Bi, Pb, Cu, Fe, and S) and the chemical communication between different minerals at grain boundaries, because addition of multiple components to a melt system generally lowers the eutectic of that system (e.g., Frost et al., 2002).

Complete melting of the polymetallic inclusions in air or in an inert atmosphere was not observed at temperatures as high as 450 °C in the Linkam heating-freezing cell. Nevertheless, complete melting of the inclusions was observed in a hydrothermal diamond anvil cell between 360 and 396 °C (Fig. 4; Fig. S3), as shown by the transformation of the melt inclusions into spherical droplets. These temperatures overlap with the total homogenization temperatures of the coexisting fluid inclusions between 293 and 400 °C (Table S2), suggesting that the now-crystalline polymetallic inclusions were trapped as melt droplets.

The internal textures of the polymetallic inclusions also indicate that they were trapped as liquid droplets. The polymetallic inclusions invariably contain suites of multiple mineral phases that extend down to the nanoscale and display a consistent crystallization sequence (Figs. 1 and 2; Fig. S1), from early to late: chalcopyrite, tellurobismuthite–rucklidgeite–native gold, calaverite, altaite, and petzite. Chalcopyrite was the earliest phase to crystallize, growing from inclusion walls with intergranular spaces filled by later phases. Native gold, tellurobismuthite, and rucklidgeite crystallizes after chalcopyrite and before other tellurides. The three minerals display undisrupted crystal faces against other tellurides. Petzite always occurs as anhedral grains and displays low grain-boundary angles against adjoining phases, suggesting it was the last phase to crystallize and thus fills any remaining space in the inclusions.

Quartz is commonly regarded as a chemically inert mineral. However, surface defect sites on quartz are very reactive, and these surface defects possess a high capacity to adsorb and reduce metals (Heinhorst and Lehmann, 1994; Mukherjee et al., 2002; Mercadal et al., 2021). When silicate minerals are mechano-chemically activated, the atomic bonds of SiO2 are broken. Reactive sites including ≡Si and ≡SiO radicals and ≡Si+ and ≡SiO ions are created. These species can recombine with each other to form siloxane bonds (Si─O─Si) or react with H2O molecules to form silanol groups (≡Si─OH) and hydrogen radicals (Kita et al., 1982):


The hydrogen radicals then recombine to form H2 molecules. This process has been experimentally proven by crushing quartz under water-saturated conditions (Kita et al., 1982). High-velocity friction experiments on various rock types, simulating earthquakes, have reproduced the generation of hydrogen as a linear function of frictional work, i.e., H2 generation increases with earthquake magnitude following a power-law relationship (Hirose et al., 2011, 2012). Frictional work at elevated temperature to >~400 °C leads to the formation of very fine-grained reactive materials at the nanometer scale; free radicals on the fresh surfaces of the fine-grained particles react with H2O, leading to the generation of H2 (Hirose et al., 2011). Strong H2 enrichments have been reported in pseudotachylites formed by fracturing on fault planes (McMahon et al., 2016) and in active earthquake zones such as the San Andreas fault (California, USA; Wiersberg and Erzinger, 2008). Accordingly, we propose that Au and, by extension, also Ag, Bi, and Te metal complexes can be reduced by silanol groups (Mukherjee et al., 2002; Hofmeister et al., 2002) via reactions such as:


or by H2 molecules (Merga et al., 2010; Mohammadnejad et al., 2013) via reactions such as:


Once metal atoms are fixed onto the quartz surface, the dispersed atoms tend to agglomerate into larger clusters via Ostwald ripening. Importantly, in our case of a Au-Ag-Te-Bi–rich assemblage at 300–400 °C, the assembled atom clusters do not form critical nuclei that subsequently grow out into solids. Instead, these clusters grow as liquid droplets because their melting temperatures are lower than the fluid temperature.

Polymetallic melt droplets, once formed, can catalyze the decomposition of metal complexes at the liquid metal-solution interface and subsequently absorb metals from solution (e.g., Trentler et al., 1995; Daeneke et al., 2018; Jian et al., 2021). Given that formation and growth of polymetallic melts are essentially adsorption-reduction reactions that do not require fluid saturation with respect to the constituent metals (e.g., Widler and Seward, 2002), this multistage process provides a mechanism by which solution components far below bulk saturation can be efficiently scavenged. For instance, the partition coefficient for Au between an aqueous fluid and bismuth melts is of the order of 107 for conditions typical of orogenic gold deposits (300 °C, pH 5; Tooth et al., 2008).

Regarding quartz, a high density of surface defects is essential for the adsorption-reduction of gold. For instance, silica with a high density of surface defects (i.e., mesoporous silica, laser-irradiated quartz surfaces) is commonly used to synthesize gold nanoparticles (Kan et al., 2003; Mercadal et al., 2021) through the adsorption-reduction of gold complexes from solution on silica surfaces. Quartz is also known for its preg-robbing behavior in gold processing (Mohammadnejad et al., 2014). Fine-grained quartz (0.1–2.5 μm) can adsorb 98% of dissolved gold from solution in hours through adsorption-reduction of gold from solution, and grinding of quartz can considerably increase its adsorption potential due to increased surface area and physical defects (Mohammadnejad et al., 2013, 2014). Therefore, intense fracturing of quartz, such as in large shear zones, now partially recorded by densely distributed secondary fluid inclusion planes (i.e., healed micro-fractures), creates permeability and a high density of surface defects and, thus, favorable conditions for the adsorption-reduction of gold and, by extension, also Ag, Te, and Bi. Gold has the highest electronegativity (2.54; Pauling, 1960) among all metals; therefore, positive gold ions should generally be more easily reduced than those of other metals. However, further studies are required to fully understand the mechanism leading to the selective enrichment of Au-Ag-Te-Bi in polymetallic melt droplets, such as studies on metal complex species and their redox potentials, investigation of properties of polymetallic metal droplets, and experiments to replicate the formation of polymetallic melt droplets through quartz fracturing under ore-forming conditions.

Lastly, the low melting point of the Au-Ag-Te-Bi–rich melt suggests that polymetallic melts could remain molten or partially molten for a long time. Thus, given favorable conditions, relatively low volumes of polymetallic melts can continue to scavenge gold from multiple pulses of ore fluids that are commonly involved in the formation of large gold deposits (Jian et al., 2022, 2024). For instance, polymetallic melt droplets previously trapped in quartz could be released due to quartz fracturing and again be exposed to aqueous fluids.

In summary, we propose a model of ultrahigh-grade gold enrichment through quartz fracturing, adsorption-reduction of metals on reactive mineral surfaces, and formation and growth of polymetallic melt droplets, which catalyze the decomposition of metal complexes and scavenge gold during protracted shearing and fluid migration. Gold particles themselves are also known to act as catalysts during the reduction of positive gold ions (Polte, 2015). Repeated fracturing of quartz could then possibly also trigger self-catalyzed growth of metallic gold without the mediation of polymetallic melt droplets.

1Supplemental Material. Methods S1; Tables S1–S4; Figures S1–S3. Please visit to access the supplemental material; contact with any questions.

This research was jointly funded by the National Natural Science Foundation of China (419720932) and the Fundamental Research Funds for the Central Universities (China; 2652020026). Bin Shi is thanked for his assistance with the SEM analysis. Qiang Liu is thanked for his assistance with the hydrothermal diamond anvil cell experiments. Li Jiang, Yongqi Su, Jinfeng Li, and Jinyu Shi are thanked for drawing the illustrations. Constructive reviews by Andy Tomkins and Krister Sundblad considerably improved the paper and are greatly acknowledged.

Gold Open Access: This paper is published under the terms of the CC-BY license.