Boiling-induced formation of colloidal gold in black smoker hydrothermal fluids

Boiling has long been recognized as an effective mechanism for depositing precious metals from solution, including as scales in geothermal wells (Drummond and Ohmoto, 1985; Brown, 1986; Spycher and Reed, 1989; Clark and Williams-Jones, 1990; Simmons and Browne, 2000). In terrestrial epithermal deposits, where boiling may also contribute to metal loss from solution (Simmons et al., 2016), gold, silver, and other elements coprecipitate with colloidal sulfides, related to the transport of gold as sulfide species in these fluids (Pope et al., 2005). However, gold-depositing processes, specifically the precipitation of discrete gold colloids, have not been observed directly. The first experimental evidence for the stability of gold colloids in hydrothermal systems was obtained nearly 80 years ago (Frondel, 1938), but there has been little agreement about the role of these colloids in ore formation (Saunders, 1990; Herrington and Wilkinson, 1993; Saunders and Schoenly, 1995; Williams-Jones et al., 2009; Saunders et al., 2010, 2014). Instead, experimental studies of gold in high-temperature hydrothermal fluids have focused on aqueous species, with sulfide as the major ligand complexing Au in solution under conditions typical of black smoker fluids similar to those considered here (e.g., Benning and Seward, 1996; Stefánsson and Seward, 2004). Under such conditions, cooling and oxidation occur as a result of hydrothermal fluid–seawater mixing and may result in gold precipitation, sometimes producing native gold in hydrothermal chimneys (Herzig et al., 1993; Hannington et al., 1995; Moss and Scott, 2001, Törmänen and Koski, 2006). In this paper we describe the first occurrence of colloidal gold in fluids from actively boiling hydrothermal vents in the Lau Basin (South Pacific Ocean). We present evidence that gold colloids form as a result of boiling, and that a fraction of the colloidal gold escapes deposition at the boiling horizon and continues to be transported with the rising fluids, explaining the high concentrations of gold in all of the chimney samples and suggesting that colloids may be an important transport mechanism for gold throughout aqueous ore-forming systems.

The sampled vents occur on Niua South, the southern of two volcanic cones that make up the Niua Volcano, at the northern termination of the Tonga volcanic arc (15.164°S, 173.757°W; Fig. DR1 in the GSA Data Repository¹; Resing et al., 2011). High-temperature hydrothermal venting is restricted to the floor of a 500-m-diameter crater on the eastern flank of Niua South, where dozens of active vents occur within an area of 150 × 200 m. The vents compose low mounds of massive sulfide capped by 1–5-m-tall sulfide chimneys composed mainly of pyrite, sphalerite, chalcopyrite, barite, and anhydrite. The deepest black smoker vents occur at a depth of 1164 m, have a maximum recorded temperature of 325 °C, and are actively boiling (Fig. 1A).

Hydrothermal fluid samples were collected from six sites by the remotely operated vehicle (ROV) ROPOS, using titanium major samplers (750 mL). Prior to sampling, the temperature of the orifice was measured using the high-temperature probe on the ROV. Four of the six sites were sampled in duplicate. Samples were processed immediately upon shipboard recovery, at which point the temperature and pH of the fluids were also recorded. Separate aliquots of the whole (i.e., unfiltered) fluid were taken for measurement of major and trace elements, as well as acid volatile sulfide and chromium reducible sulfide. Because the fluids were unfiltered, larger particles (>0.2 µm) may also have been present in the fluids. For major elements, samples were acidified with HCl to below pH 1 and were further leached upon return to shore with nitric acid prior to analysis by inductively coupled plasma–mass spectrometry.

Following subsampling of the liquid, the remaining fluid was spun down using a shipboard centrifuge (Gartman et al., 2014). The resulting pellet was then capped with N₂ and frozen at −20 °C. In the laboratory, the frozen pellets were resuspended in Mill-Q water and evaporated directly onto aluminum stubs for scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM/EDS) analysis on a Tescan VP-SEM in high vacuum mode, without conductive coating. Imaging was performed using both secondary electron and backscatter detectors, and particle dimensions were measured using ImageJ (https://imagej.net/) software. Samples of the chimney walls collected at the vent orifice were examined using an FEI (https://www.fei.com) MLA (mineral liberation analyzer) 650F SEM under high vacuum and also analyzed for bulk Au content by instrumental neutron activation analysis. Additional details of the sampling and analytical methods used for the fluids, particles, and chimney samples are provided in the Data Repository.

Gold particles ranging in size from <50 nm to 2 µm were found in both focused and diffuse flow fluids at Niua, as well as on a black tarnish, mainly composed of metal sulfide minerals, that developed on the exterior of the snorkel during sampling of the boiling vent (Fig. 1; Fig. DR2). The majority of particles have anhedral shapes ranging from round (n = 18) to oblong (n = 14), with average dimensions of 0.44 × 0.31 µm (median 0.33 × 0.20 µm). Almost all (95%) particles had at least one dimension that was shorter than 1 µm; 29% of the particles exhibit a dimension of <100 nm and can be considered nanoparticles (Fig. 2A). The majority of larger (>500 nm) particles appear discrete and likely result from particle growth (Fig. 2B; Fig. DR3), although apparent colloidal aggregates are also present (Fig. 2A). Chalcopyrite and barite were also present in the fluids, and two of the gold particles are embedded in barite. The majority of Au particles appear to be native Au with Ag levels below the detection limits of EDS, with the exception of 4 particles containing Ag between 4.6 and 6.9 mol% (2.5–3.8 wt%) relative to Au.

Gold particles were also found in sulfides from the wall of the diffusely venting chimney; two of the particles are <1 µm in size, the other four are larger (2–5 µm) than any particles found in the fluids (Figs. 2C and 3). Chimney samples from the focused flow and boiling vents contain an average of 6 mg/kg of Au (ranging from 3.6 to 11.4 mg/kg, n = 5; Table DR1). A sample from the diffusely venting chimney contains >30 mg/kg Au. Like the Au in the fluid, these particles are associated with chalcopyrite and barite.

Gold concentrations in the fluid samples (unfiltered fluids, which include dissolved and particulate gold) range from 3.1 to 3.7 nM in the boiling vents, and 1.6–5.4 nM in the focused flow vents (Table DR1). The number of gold particles observed in the samples from the focused flow fluids (n = 6, 3, and 1) correlates generally with the bulk gold concentrations (3.1 nM, 2.0 nM, and 1.6 nM, respectively; Fig. 4); mass balance calculations relating the amount of gold present in particles to the amount of gold measured in fluids suggest that most of the gold is present as colloids. The diffuse fluid sample, which contains the most (n = 25) and the largest gold particles, has the highest bulk Au concentration (26.7 nM) of the analyzed samples.

The highest temperature sample was both visibly boiling and plots directly on the seawater boiling curve (Fig. DR4). At Niua, the solubility of gold under pre-boiling conditions is ∼10× higher than the concentration in focused flow fluids that were collected after boiling (Fig. DR5). However, sulfide loss at the point of boiling due to both vaporization and the observed precipitation of metal sulfides could have resulted in fluid supersaturation with gold carried originally as either Au(HS)₂^– and AuHS⁰, resulting in the rapid nucleation of colloidal gold particles (e.g., Saunders and Schoenly, 1995). The concentrations of sulfide measured here are low (Table DR1), consistent with loss upon boiling. Gold concentrations in the vent fluids at Niua do not correlate with Mg, a measure of seawater–hydrothermal fluid mixing, consistent with the gold transport in fluids being dominated by particles rather than dissolved species.

Deposition of colloidal gold at the point of boiling was observed on the black tarnish that precipitated on the sampler snorkel (Fig. DR2), and high concentrations of gold were present in the corresponding chimney sample, although no gold particles were observed in the chimney or the boiling fluids. It is probable that the initial gold colloids that form upon boiling are smaller than we could observe with SEM, as initial gold nanoparticles precipitated via common boiling syntheses often range from 1 to 20 nm (Daniel and Astruc, 2004). The particles that occur in the focused flow fluids, diffuse fluids, and chimney wall, in which irregularity of shape increases with size, are consistent with particle growth after boiling-induced nucleation (Figs. 3 and 4).

Indirect evidence of solid precious metal nanoparticles in fluid inclusions from epithermal veins (Kouzmanov et al., 2010) and isotopic evidence (Saunders et al., 2016) present the possibility that in some systems, gold particles may exist in the fluids even prior to boiling. This process may be particularly common at arc volcanoes where hydrothermal fluids are pre-enriched in gold and other elements as a result of magmatic volatile input (Hedenquist and Lowenstern, 1994; de Ronde et al., 2011; Berkenbosch et al., 2012; Schmidt et al., 2017). However, at Niua, the formation of gold particles on the snorkel tarnish, the elevated concentration of gold in the boiled chimney, the undersaturation of Au in pre-boiled fluids (Fig. DR5), and the transport of particulate gold in boiling, focused, and diffuse flow fluids, suggest that the precipitation at Niua is boiling induced, as previously suggested for bonanza Au-Ag precious metal vein mineralization at the Sleeper deposit in Nevada (Saunders and Schoenly, 1995) and in gold-supersaturated seawater-dominated fluids in the Reykjanes (Iceland) geothermal field (Hannington et al., 2016).

The occurrence of gold colloids in hydrothermal fluids of the Niua vent field indicates a process by which gold is initially precipitated in high-temperature boiling fluids and subsequently transported in particulate form into lower temperature vents (Fig. 4). At depth, before the hydrothermal fluids reach the seafloor, gold is most likely present in solution, as suggested by solubility calculations (Fig. DR5). Boiling, and the rapid loss of sulfide to exsolution and to the formation of metal sulfides, induces gold precipitation as colloids. Some of the colloidal gold adheres to local chimney walls, and the remainder is carried upward in suspension, potentially stabilized by capping agents present in hydrothermal fluids (e.g., thiols; Reeves et al., 2014). As the fluids cool (Fig. 4), the colloids aggregate and grow, with some additional loss from suspension. At the lowest temperature site, the fluid is diffusely vented, resulting in particle settling that is more rapid than particle outflow and concentration of the gold colloids and particles within the diffusely venting fluids. Diffuse venting in white smokers has long been linked to gold enrichment due to mixing and cooling of the hydrothermal fluids and associated loss of aqueous sulfur complexes (Hannington et al., 1986). The observations at Niua suggest an important alternative mechanism for gold transport into the low-temperature vents.

The presence of colloidal gold particles in hydrothermal fluids and sulfide chimneys at Niua South has several implications for the formation of Au-rich ore deposits: it confirms a nearly 80-yr-old hypothesis that Au can be transported colloidally in geothermal solutions, further demonstrates that Au colloids can remain suspended after boiling, and shows that the accumulation of colloids in boiled liquids can eventually result in gold concentrations that exceed saturation values. These findings provide direct observational insight into the role that colloids may play in high-grade gold mineralization, especially at seafloor massive sulfide deposits.

We thank the Schmidt Ocean Institute (Palo Alto, California, USA) for providing ship time, as well as the crew of the R/V Falkor and the ROV ROPOS for their skilled support. We thank the Kingdom of Tonga for permitting this work within their territorial waters. Gartman thanks the U.S. Geological Survey Mendenhall Research Fellowship Program for funding. Findlay acknowledges funding from a Fulbright Postdoctoral Research Fellowship. Kwasnitschka acknowledges support of general cruise logistics through the Helmholtz Alliance ROBEX. Financial support for this work was provided by GEOMAR–Helmholtz Centre for Ocean Research Kiel (Germany) and a Natural Sciences and Engineering Research Council of Canada Discovery Grant to Hannington and a DFG Major Research Instrumentation Programme grant to Garbe-Schönberg at Kiel University. Jamieson and Peterkin thank Graham Layne for insightful discussions on gold precipitation and scanning electron microscope analysis, and acknowledge funding from the Canada Research Chair Program and a Newfoundland RDC Ignite grant. A subsidy for mobilization of ROPOS provided by the Canadian Scientific Submersible Facility was enabled by the Canada Foundation for Innovation Major Science Initiatives Fund (CFI MSI).