Abstract
Many exceptionally Au-rich deposits are associated with subduction zones and Au-rich magmas. However, the origin of Au enrichment in arc magmas remains controversial. Here, we present a new high-precision geochemical dataset on 66 submarine glasses from the length of the oceanic Kermadec arc and Havre Trough back-arc system. The combined systematics of Au with other chalcophile elements (Ag, Cu, Se, Pt) shows that hydrous melting in the Kermadec subarc mantle occurs predominantely at relatively high temperatures above the sulfide liquidus and produces magmas with mantle-like Ag/Cu values. The highest Au (up to 6 ng/g) and Au/Cu (up to 0.06 × 10-3) exceeds that in fertile mantle and primitive MORB by far, requiring remelting of previously depleted mantle. We demonstrate that high-degree multiple-stage melting of hydrous and oxidized subarc mantle are major drivers of Au enrichment of parental magmas in intra-oceanic island arcs, such as the Kermadec arc.
Similar content being viewed by others
Introduction
Some of the highest-grade Au deposits associated with Cu- and Ag-rich mineralization (i.e., volcanic hosted massive sulfides) are found above subduction zones1,2. At the ocean floor, arc and back-arc deposits make up 40% of seafloor massive sulfide occurrences globally, which can form exceptionally high-grade deposits (with up to 43 g/t Au) compared to those associated with mid-ocean ridges3. Sparse data on Au contents in basaltic magmas generated in oceanic subduction zones indicate that these magmas are enriched in Au in comparison to mid-ocean ridge basalts, suggesting a possible genetic link between parental magma compositions and Au endowment of ore deposits4,5,6,7,8.
Gold has a strong affinity for sulfide phases during mantle melting and subsequent magma fractionation e.g., refs. 9,10,11. In the absence of sulfide, Au behaves as an incompatible element with respect to common silicate phases and is concentrated in silicate melts12. Many studies have recognized a key role for prolonged sulfide undersaturated crystallization in generating magmas enriched in Au and other strongly chalcophile elements, such as Cu and Ag e.g., refs. 7,11. However, it remains controversial whether parental subduction-related magmas are more enriched in Au than, for example, parental mid-ocean ridge basalts, and if so, what the major factors are that control this enrichment.
The extent to which sulfide is consumed during mantle melting, releasing Au to the magma, depends on the degree of melting, the sulfur solubility in the melt at given physical conditions (P, T, ƒO2) in the mantle, and on the melt composition7,13,14. Variations in the degree of partial melting and melting conditions may therefore represent some of the dominant factors controlling the magmatic Au, as well as Ag and Cu inventory15,16,17,18. Additional factors controlling the Au endowment of parental mantle magmas could be variations in the subarc mantle depletion due to previous episodes of melting9,18,19,20, input of Au-bearing fluids/magmas from the subducting plate21,22 and also recycling sulfides in the crust23.
To date, there have been several case studies of Au behavior in Western Pacific arc systems focused on deciphering the role of sulfide or fluid phases in Au concentration and transport in evolved magmas7,24,25,26. Unlike these studies, the main focus of this work is to analyse the possible initial diversity of arc magmas in terms of Au concentrations, evaluated on a large regional scale. Pillow rim glasses from the intra-oceanic Kermadec arc studied here span the length of the arc, are compositionally diverse and were formed under variable geodynamic conditions ranging from the subduction of thin Pacific oceanic crust at the northern segment to subduction of thick Hikurangi oceanic plateau crust beneath the southern segment. Furthermore, the amount of subducted, New-Zealand-derived sediment decreases northwards along the arc. The Kermadec arc is, in many respects, a unique setting for studying the processes occurring in the sub-arc mantle and during early magmatic fractionation in the crust that influence the Au budget beneath intraoceanic arcs.
Results and discussion
Geological background
The intra-oceanic Kermadec arc and Havre Trough comprise a ~ 1300 km-long arc-back-arc system located between New Zealand and Tonga (~25 °S to ~36 °S; Fig. 1). Dehydration of the Pacific Plate, subducting westward beneath the Australian Plate, drives arc magmatism and the formation of arc front volcanoes, most of which host active hydrothermal systems and also some massive sulfide mineralization with high grades of Au2,27,28. Submersible dives have been conducted on 14 hydrothermally-active volcanoes of the Kermadec Arc with four of the volcanoes (29%) found to host sites with massive sulfide mineralization. Some of the best studied are Brothers and Clark Seamounts28,29. At Brothers Volcano, the main zone of mineralization extends over an area of at least 600 × 200 m with alteration measured (by magnetics) to ~450 m below the seafloor with Au grades of up to 90 ppm occuring in some chimneys30. The majority of the Kermadec arc volcanoes, however, have not been surveyed in comparable detail, thus the number that host massive sulfides may be higher.
Slab surface model is from the United States Geological Survey80, derived from relocated seismic events and from ocean bottom seismometer wide-angle refraction seismic data. Inset shows a globe with the location of the Kermadec arc. Symbols mark sampled locations of Kermadec arc front volcanoes and Havre Trough back-arc. Gray lines mark slab depth contours spaced at 20 km. Arrows with numbers are subduction velocity rates after ref. 81.
At the Kermadec trench, the incoming Pacific Plate changes markedly in thickness and composition along strike, with the Hikurangi Plateau –a ~ 9-15 km thick Cretaceous LIP – currently subducting south of ~34.5 °S27. To the north of this latitude, ~6 km thick typical oceanic crust comprises the uppermost part of the subducting Cretaceous-age Pacific Plate. South of 34.5 °S, where the oceanic Hikurangi Plateau subducts beneath the southern Kermadec arc, the total volume of erupted lava is the greatest (~1200 km3, compared with ~830 km3 north of 34.5 °S), and the volcano spacing is lowest (~30 km, compared to ~55 km to the north)2. Furthermore, the slab surface models show that the angle of the subducting Pacific Plate changes from ~30 ° at ~60 km depth to 47 ° at ~100 km depth beneath the northern Kermadec arc compared to 20 ° at ~60 km depth to 56 ° at ~100 km depth beneath the southern Kermadec arc (Fig. 1). The Pacific slab surface beneath the arc front volcanoes, however, remains relatively constant between ~100 and 120 km depth. Lavas from the Kermadec Arc show a marked increase in radiogenic Pb, and to a lesser extent in 87Sr/86Sr, and a decrease in 143Nd/144Nd isotope ratios south of ca. 32 °S, interpreted as increased material flux from the subducting Hikurangi Plateau and/or sediments derived from New Zealand31,32 or a different ambient mantle composition e.g., ref. 33.
Volcanic glass composition
Volcanic glasses analyzed in this contribution have been collected during seven surveys over the last 20 years with R/Vs Tangaroa (NZAPLUME I-III (1999-2004); NZASMS (2011); NIRVANA (2012); Ka-imikai-o-Kanaloa (NZASRoF’05 in 2005) and Sonne (SO255 Vitiaz in 2017) using rock dredges and the submersible Pisces. In total 56 volcanic glass samples have been retrieved from 17 volcanoes from pillow-rims and scoria along the length of the Kermadec arc located between 36.4 °S (Clark volcano) and 25.8 °S (Monowai). Another 10 pillow-rim glasses have been sampled from the Havre Trough back-arc between 34.4 °S and 28.1 °S. Sample locations and water depth for each sample are listed in Supplementary Data 1. For comparison we also present here new data on various MORB glasses, mostly from the Atlantic and Arctic oceans (cf., Supplementary Data 2). These data were obtained using the same analytical technique as the Kermadec data set and are thus fully consistent with the latter (see analytical methods).
The compositions of the Kermadec arc glasses range from basaltic to rhyolitic and fall into the low-K tholeiitic to medium-K calc-alkaline series (see Supplementary Data 1; Supplementary Fig. 1). Major element systematics are consistent with predominant olivine + pyroxene ± plagioclase crystallization at MgO >4 wt% (Fig. 2A–B; Supplementary Fig. 1), where contents of SiO2, FeOt, Na2O, TiO2 increase, CaO and Al2O3 decrease with decreasing MgO. In the MgO range of 3–4 wt%, FeOt, TiO2 begin to decrease and SiO2 increases more sharply with decreasing MgO, suggesting the onset of magnetite ( ± ilmenite) crystallization6 (Fig. 2A–B).
Kermadec arc literature data are from refs. 38,39,47,40,41,82. Compositions of mid-ocean-ridge basalt glasses are shown after83 and this study. Depleted MORB melt (DMM) composition is after69. Blue lines mark schematic sulfide fractionation trajectories (SL = sulfide liquid; MSS = mono-sulfide solid solution). Gray field marks the onset of magnetite (mt) and sulfide (sulf) fractionation. Diagram panels show wt.% MgO versus (A) wt.% SiO2, (B) wt.% FeOt, (C) µg/g Se, (D) ng/g Au, (E) µg/g Cu, (F) ng/g Ag, (G) 1000x (Ag/Cu) and (H) 1000x(Au/Cu).
Gold contents in the Kermadec glasses range from <1to 15 ng/g (Fig. 2D). With decreasing MgO from 8 to ~5 wt%, the maximum Au contents and their range increase and then decrease at lower MgO. Selenium, which in contrast to sulfur is not volatile at pressures of ≤100 bar34, and Cu ( ± Ag), demonstrates qualitatively similar behavior (Fig. 2C and 2E-2F). The decreasing Au, Cu and Se contents at MgO ≤4 wt% are consistent with sulfide saturation and strong partitioning of these elements into sulfide phases. It is noticeable that sulfide saturation coincides closely with the Fe-Ti oxide saturation and could be triggered by the latter e.g., ref. 7.
A strong control of a sulfide phase on the composition of dacitic to rhyolitic Kermadec glasses is also evident from their very high Au/Cu and particularly Ag/Cu values (Figs. 2G and 2H), similarly to what has been described previously in Brothers volcano, Manus and Lau Basin magmatic suites7,8,26. This fractionation can be explained by crystallization of a Fe-Ni monosulfide solid solution ((Fe,Ni)1-xS; MSS) phase, in which Cu is more compatible element than Au and Ag14,26. Alternatively, such inter-element fractionation can be explained by assimilation of Au- and Ag-rich and to a lesser extent Cu-rich intermediate solid solution (ISS) from a previously formed sulfides in cumulative horizons of magma chambers23. This process of metal concentration in a sulfide phase or silicate melt is an important pre-conditioning for formation of Au-Cu-Ag-rich hydrothermal sulfide deposits (e.g., refs. 4,23,34,35). Much less is, however, known about the earlier stages of arc magma fractionation and the factors governing the noble metal endowment of primitive magmas, which is the focus of this study.
Gold systematics in primitive Kermadec melts
Primitive sulfide-undersaturated Kermadec glasses span a significant range in MgO contents (4–8 wt%), and thus their compositions are affected by variable extents of fractional crystallization leading to a variable enrichment in Au and other chalcophile elements. In order to compare glasses with different MgO, we normalized the contents of chalcophile elements to MgO = 8 wt% using the equation proposed by Taylor and Martinez (2003)36 for incompatible elements in back-arc basin glasses. For example, the equation for Au has the following expression:
\({{{\rm{Au}}}}_{8}={{{\rm{Au}}}}\times {({{{\rm{MgO}}}})}^{1.7}/34.3\)
As illustrated in Fig. 3, Au8 contents in primitive Kermadec arc glasses range from 1-8 ng/g. Most glasses are more enriched than primitive MORB with MgO ≥ 8 wt%, which contain ≤ 2 ng/g Au37,38, this study. Except for one sample from Giggenbach volcano with Au8 > 4 ng/g, volcanic glasses from the northern Kermadec arc segment are the most Au-rich. Glasses from the southern segment are systematically less enriched and have Au8 < 4 ng/g. Glasses collected between 31 and 34.5oS and also Havre Trough glasses have intermediate compositions and range from MORB-like with sub-ng/g Au8 contents to 6 ng/g. Havre Trough glasses reveal a crude inverse correlation of Au8 with the distance to the volcanic front.
Covariations of Au8 vs (A) Cu8, B Pt8, C Se8, D Na8, E Nb/Zr and F Zr/Hf in mafic Kermadec glasses (8 = corrected to MgO = 8 wt%). Gray rectangles represent mid ocean ridge basalt (MORB) glass compositions with MgO >6 wt% are shown for comparison (data from this study). Depleted MORB mantle (DMM) compositions are after ref. 69.
Au8 correlate positively with Cu8, Se8, Pt8, (Fig. 3A–C). Remarkably, high Au8 correspond to high Au/Cu (up to 0.06 × 10−3), significantly exceeding typical mantle ratios of ~0.03 × 10−3 (Fig. 2H). In contrast, Ag/Cu do not shown large variations in the mafic Kermadec glasses and, with a few exceptions, remain in the mantle and MORB compositional range of Ag/Cu=0.20-0.45 × 10−3 (Fig. 2G). Exceptionally high Ag/Cu = 0.6 × 10−3 were found only in high-Mg andesite glasses from Kibblewhite39,40,41. When plotted against incompatible lithophile trace elements that are proxies of mantle source depletion and degree of melting, Au8 correlate negatively with Na8 and ratios of incompatible lithophile elements (Nb/Zr, and Zr/Hf; Fig. 3D–F). Correlations of predominantly slab-derived elements (e.g., Cl, U, Pb and Ba) and proxies of slab-derived components such as B/La (not shown), Ba/Th or Th/La tend to have positive slopes, but they are statistically less significant than correlations with proxies of mantle depletion (Supplementary Fig. 2).
Possible crustal control on gold enrichment
Gold enrichment observed in the mafic Kermadec glasses can be related to mantle melting in the presence of slab-derived fluids or melts, or it may have been modulated by processes occurring during the earliest stages of magmatic fractionation, including potential interaction of primitive magmas with wall rocks during ascent. In particular, a model to generate Au-rich primitive arc magmas in the lower crust was proposed by Holwell et al.23. This model suggests that ascending arc magmas can mobilize Cu–Au-rich sulfide melts produced by partial melting of previously formed sulfides in lower-crustal cumulates. This process is expected to enrich the melts in a number of chalcophile elements (Cu, Au, Ag, Pt, Te) that have a stronger yet variable affinity for Cu-rich sulfide liquid compared to the more refractory MSS.
Because of the variable affinity of elements for sulfide liquid and MSS, the process of partial sufide melting is known to strongly fractionate some chalcophile element ratios between these sulfide phases. For example, Ag/Cu and Au/Cu are ~10 times higher in sulfide liquid than in equilibrium MSS (refs. 13,14). Therefore, one would expect Au–Cu-rich sulfide liquids, and the melts that assimilated them, to exhibit elevated Ag/Cu and Au/Cu ratios correlating with Au enrichment. However, the compositions of the Kermadec glasses show that Au enrichment does not correlate with the Ag/Cu ratio, which mostly corresponds to typical MORB and mantle values (e.g., Fig. 2). Moreover, the model of Holwell et al. (2022) does not predict a correlation between Au enrichment and incompatible lithophile elements (Fig. 3D–F), which we observe in the Kermadec magmas. Therefore, we consider this process of Cu–Au-rich sulfide mobilization in the lower crust unlikely to operate on a regional scale or to govern the systematics of the Kermadec glasses. It appears more likely that the variations in Au enrichment were generated during mantle wedge melting.
Is addition of gold to the subarc mantle source required?
Gold can be highly soluble in aqueous fluids by forming complexes with a range of bisulfide, chloride and polysulfide ions21,42,43,44. For example, the trisulfur radical S3-, to which Au has a strong affinity, is stable in aqueous fluids up to ~700 °C and ~3.5 GPa20,45. Some of the Au in the sub-arc mantle may therefore have been added by aqueous, S-Cl bearing fluids from the subducting slab. For example, He et al.46 proposed that aqueous slab-derived fluids are responsible for Au enrichment in the mantle wedge, its oxidation and triggering partial meting, which is a prerequisite for formation of Au-rich deposits in subduction-related settings. This view is testable using our data from Kermadec arc as these melts undoubtedly originate from a hydrated and oxidized subarc mantle5,33,39,40,47.
Slab-related Au enrichment in hydrous arc lavas is hypothesized on the observational basis that these lavas are enriched in Au compared to relatively “dry” mantle derived magmas, from mid-ocean ridges39, and that experimentally determined Au solubility is high in potential slab-derived fluids21. However, the key question remains unanswered: is Au enrichment coupled with other slab-related enrichments in arc lavas and variable subduction inputs?
To address this question, we refer to rather insignificant correlations of Au enrichment (Au8) in Kermadec glasses with absolute contents of mobile in hydrous fluids elements (Cl, U, Ba), or their ratios over less mobile elements (e.g., Ba/Th; cf. Supplementary Fig. 2). An exception is the relatively good correlation between Au8 and Pb8, which could also be related to a major control of sulfide phases on the content of both elements. Thus, there is no unequivocal evidence for coupled behavior of Au and elements mobile in slab-derived fluids, which would support the largely fluid-mediated mobility of Au in the mantle wedge beneath the Kermadec arc.
The highest Au contents were measured in glasses from the northern portion of the Kermadec arc, where the subducting crust and mantle have MORB-type composition33,48 covered by <200 m of sediments49. MORB and most sediments typically contain low Au contents of ≤1 ng/g38. In contrast glasses from the the southern segment, where Hikurangi Plateau is subducting under the Kermadec arc (Fig. 1) contain less Au. Although there are no data for Au contents in the Hikurangi plateau basalts, we can expect that they are relatively enriched compared to MORB by analogy with other oceanic igneous plateaus containing an average ~4 ng/g Au50. Thus, a correlation between Au contents in the subducting plate and Kermadec arc lavas is uncertain. In contrast, incompatible trace elements and their isotope ratios show a clear spatial zonation correlating with subduction input33.
The lack of clear correlations with fluid-mobile elements and moderately high Au contents in primitive Kermadec magmas suggest rather minor subduction-related Au enrichment in their mantle source. Though, we cannot exclude that slab-derived fluids transport some Au locally to the mantle wedge as suggested by previous studies, this amount should be small compared to the mantle endowment.
How much gold in magma can be produced by mantle melting?
As Au enrichment appears not to correlate with slab-related enrichment, the question arises whether mantle melting alone can account for the observed enrichment in mafic Kermadec melts. To address this question, we modeled mantle melting under different starting conditions and compared the expected Au contents in primary magmas with those observed in the Kermadec glasses (Fig. 4). Assuming that primary Kermadec magmas have up to 18 wt% MgO in equilibrium with olivine Fo93-94, MgO = 8 wt% in evolved glasses can result from ~20-25% olivine crystallization as suggested by forward modeling in Petrolog software51 (see Methods). The maximum initial Au content in primary Kermadec magmas can be estimated as 6 ng/g from the maximum Au8 = 8 ng/g assuming perfectly incompatible behavior of Au in olivine.
A Effect of melt pooling and temperature. At high temperature, the stable sulfide phase is sulfide liquid (SL), at low temperature mono-sulfide solid solution (MSS). Au is highly compatible in SL, and low degree melts are depleted in Au until complete sulfide exhaustion from the mantle (at 20% melting in this model). Effective pooling before eruption suppresses Au contents in erupted magmas. Instantaneous melts are very rich in Au at sulfide exhaustion. Au is significantly less compatible in MSS, and the highest Au contents are achieved at low degree melting in both pooled and instantaneous melts. B Effects of S content and its solubility on Au content in mantle melts are counteracting. At higher is S solubility (for example, at mantle oxidation by slab fluids53,55) the more sulfide is melted and the more Au is released into melt. Increasing the amount of S in the mantle has the opposite effect as it increases the amount of sulfide and consequently suppresses Au content in magma. C Temperature regime and the presence of sulfide as SL or MSS has a strong effect on relative fractionation of Au, Ag and Cu. At low temperature, Au and Ag are less compatible in MSS than Cu and their mantle normalized ratios in magmas are above unity. At high temperature, Au is more compatible in SL while Ag and Cu are less compatible and have similar affinity to SL. Thus, Au/Cu ratios in magmas are low, whereas Ag/Cu ratios are close to the mantle value. D Effect of multi-stage melting on Au/Cu and also on absolute Au contents in melts arises from contrasting partitioning of Au and Cu into SL. Au/Cu increases in mantle residues during partial melting in the presence of SL. Remelting of residual mantle generates magmas with elevated Au/Cu. The total degrees of melting, when magmas can achieve the highest Au/Cu, are modulated by counteracting effects of S addition with slab-derived fluid to the mantle and enhanced S solubility at oxidizing conditions of arc magma generation.
The modeling for the case of high-temperature melting, when liquid sulfide is present in the mantle52, illustrates that the maximum Au content in primary magmas is achieved within 5–7% melting after exhaustion of sulfide from the mantle and largely dependent on the extent of melt pooling before eruption, sulfur solubility in melt and source composition. At sulfur solubility in melt of ~1000 µg/g and initial S content in the mantle of 200 µg/g, the maximum Au content in primary magmas is ~3 ng/g at ~28% melting (Fig. 4A). At degrees of melting <20%, Au content in primary magmas remains well below 1 ng/g. This regime of melting corresponds well to the generation of most MORBs, but cannot explain concentrations up to 6 ng/g Au estimated for the Kermadec arc primary melts. In the case of low-temperature melting, when MSS is stable in the mantle, the bulk partition coefficient is less than 1 and Au behaves as a moderately incompatible element. Its contents decrease from about 6 ng/g in low-degree melts to ~3 ng/g at 30% melting (Fig. 4A). These contents are similar to those estimated for parental Kermadec magmas; thus relatively low-degree low-temperature mantle melting in the presence of MSS appears favorable for generating Au-rich melts from a vommon MORB mantle.
In the case of incomplete melt mixing prior to eruption, high-degree Au-rich melts can be preferably sampled by erupted magmas. For example, during incremental melting with 1% step modeled here, a maximum theoretical Au content of ~18 ng/g is achieved immediately after sulfide exhaustion during partial melting (Fig. 4B). Therefore, the magmas, in which low-degrees melts are underrepresented during pooling before eruption, may have maximum Au contents between 3 and 18 ng/g. It is important that the maximum estimated Au content in Kermadec parental magmas does not exceed this predictable range of Au concentrations potentially generated by the mantle melting process.
Sulfur solubility – a proxy for ƒO2 in mantle53 – affects the amount of sulfide melted and thus the content of Au in partial melts. For example, increasing S solubility from 1000 µg/g to 2000-4000 µg/g, which is more typical for redox conditions at the sulfide-sulfate transition occurring in arc magmas54, increases the maximum Au content in partial melts up to 5–7 ng/g, which is close to the maximum Au in Kermadec melts (Fig. 4B). Even higher Au contents can result from remelting of residual mantle after previous episodes of melting. For example, up to 8 ng/g Au in melt can be achieved by remelting of residual S-depleted mantle at relatively oxidized conditions corresponding to S solubility of 4000 µg/g. However, if slab-derived fluids bring some S to the mantle wedge, as suggested by studies on S isotopes55,56,57, the effect of second-stage melting on absolute Au contents will be suppressed due to the increase of the bulk mantle/melt Au partition coefficient with increasing amounts of sulfide in the mantle.
To sum up, our modeling shows that high Au contents, overlapping with those in the Kermadec parental melts, can be generated by mantle melting processes under a range of conditions, in agreement with results of previous studies20,58,61,60,61. The contrasting conditions of generation of Au-rich magmas are (1) low-temperature low-degree mantle melting in the presence of MSS, and (2) high-temperature high-degree melting in the presence of sulfide liquid and beyond the sulfide stability (Fig. 4). Oxidation of mantle sources leading to increasing S solubility in silicate melts also helps to generate Au-rich magmas at any temperature.
Mantle melting control on gold abundance in Kermadec magmas
As detailed in the previous paragraphs, optimal thermal conditions of mantle melting are critical for generation of Au-rich magmas. The existing models propose a relatively wide interval of temperatures in the upper mantle wedge melting zone beneath volcanic arcs from 1100–1450 °C62,64,65,65 overlapping the stability fields of both MSS and liquid sulfide52. The data on pressure-temperature conditions of Kermadec magma generation are rather scarce and do not allow more precise estimates, particularly with good spatial resolution. In the absence of these data, the thermal conditions of mantle melting and the relative importance of MSS or liquid sulfide can be assessed directly from the systematics of Au with other chalcophile elements using available experimental data13,14.
At relatively low temperatures, when Fe–Ni-rich MSS is stable, partition coefficients between sulfide and mantle melts decrease in the order DCu > DAu > DAg, resulting in high Au/Cu and Ag/Cu values in mantle melts at low degrees of partial melting13,14 (see Fig. 4C). At high-temperature conditions above the sulfide liquidus (e.g., T > 1200 °C at 1.5 GPa52), partition coefficients between sulfide liquid and mantle melts are in the order DAu > DCu ≈ DAg resulting in low Au/Cu and mantle-like Ag/Cu values during initial melting. At intermediate temperatures, both solid and liquid phases are present and sulfide–silicate melt partitioning of elements depends on the relative proportion of the sulfide phases. After sulfide exhaustion Au/Cu and Ag/Cu values approach the mantle source values independent of the thermal conditions of melting. Based on the experimental data, Ag/Cu values in primitive magmas generated by low to moderately high degrees of melting are particularly useful to determine, which sulfide phase is stable in the mantle, and whether melting occurred at high or low temperatures.
Most Kermadec glasses show Ag/Cu values similar to those for mid-ocean ridge glasses and mantle values (Fig. 2G). This suggests that the majority of Kermadec magmas were formed at high temperatures in the presence of liquid sulfide or represent products of very high degrees of melting (>20–30%), when all the sulfide was melted out and the Ag/Cu ratios approached their source values (Fig. 4C). This result is consistent with published thermal models that predict rather high peak mantle temperatures ranging from 1380–1450 °C (model D80)63 beneath the Kermadec arc front.
Three high-Mg andesite glasses from Kibblewhite, however, have Ag/Cu of ~0.6 × 10−3 that are significantly higher than in the majority of other Kermadec glasses and MORB (Fig. 2G). These glasses also have a relatively high Au/Cu of ~ 0.04 × 10−3 and can be formed by low to moderate degrees of melting of relatively low-temperature mantle in the presence of MSS. This conclusion is in agreement with data by Hirai et al.40, who proposed that the Kibblewhite high magnesian andesites formed at relatively low pressure and temperature conditions (1.1–1.3 GPa, 1145–1230 °C) that partly overlap the MSS stability field52. Alternatively the Kibblewhite magmas can be formed by mixing between typical basaltic Kermadec magmas with mantle-like Ag/Cu and evolved high-Si magmas with high Ag/Cu, which fractionated MSS (Fig. 2G). As discussed by Hirai et al.40 this hypothesis is not confirmed by the absence of disequilibrium mineral assemblage in the andesites and their low-amount of phenocrysts that is not typical for a hybrid origin of the high-Mg andesites.
Although a very minor occurrence of low-degree melting in the presence of MSS seems reasonable, most of the Kermadec arc glasses have MORB-like Ag/Cu, which suggests that low-degree and low-temperature melting in the presence of MSS plays only a minor role in the Kermadec subarc mantle. Partial melting in the stability field of MSS, however, may be more important at active continental margins with thickened crust and a colder mantle wedge (e.g., in Andes), where it can be recognized based on elevated Ag/Cu values in primitive melts59.
During high-temperature melting in the presence of liquid sulfide, Au contents close to those in parental Kermadec magmas can be obtained only at high degrees of melting and/or at high S solubility in melt at moderately oxidized conditions (Fig. 3). Both these prerequisites are met in the Kermadec arc, and the validity of such a view is confirmed by correlations between Au8 and indices of mantle depletion and/or extent of melting (Na8, Nb/Zr, Zr/Hf; Fig. 3D–F). However, a simple model of extensive one-stage mantle melting fails to explain significant variations of Au/Cu in Kermadec glasses, partly exceeding mantle Au/Cu~0.03 × 10−3 values (Fig. 3), and also strong correlation of Au/Cu and Au8.
A plausible explanation for super-mantle Au/Cu (and Au/Se) values in Kermadec magmas requires the adoption of a model of multi-stage mantle melting. Preferential incorporation of Cu over Au during mantle melting in the presence of sulfide liquid results in increasingly higher Au/Cu values in the residual mantle up to the point of sulfide exhausting (Fig. 4D). Therefore, a subarc mantle that underwent previous melt extraction has higher Au/Cu than unmelted mantle. Gold to Cu values (at MORB-like Ag/Cu) in glasses from Monowai and Putoto (northern Kermadec arc) and from Rumble III, Giggenbach, Kuiwai, Kibblewhite, Ngatoroirangi and some near arc front glasses from the Havre Trough are higher than those of depleted mantle and MORB. The highest Au/Cu values of up to 0.06 × 10−3 (~2× higher than in primitive mantle) is observed in moderately mafic glasses from the northern Kermadec arc volcano Putoto. These glasses also have the lowest Na8, Nb/Zr (and Zr/Hf) suggesting their formation from the most depleted subarc mantle. Alternatively, high Au/Cu values can be generated by incomplete pooling of partial mantle melts, assuming that low-degree melts (i.e., <10%) are extracted from the mantle and, for example, solidified during ascent. A successful model of multi-stage mantle-melting explaining variable Au contents, Au/Cu and Ag/Cu values in parental Kermadec magmas is presented in Fig. 5.
All symbols are as in Fig. 3. Au in primary Kermadec magmas were calculated by reducing Au8 in primitive glasses by 20% relative that implies ~20% crystallization from primary melt to 8 wt% MgO. Different lines correspond to melting trajectories at different proportions of SL and MSS during melting, different S content in the mantle (300 or 200 µg/g), different solubility of S in melt (2000 or 4000 µg/g), reflecting less to more oxidizing conditions, and different source composition ranging from DMM to residual mantle after 10, 15 and 20% DMM melting. Primitive MORB compositions with MgO >8 wt% are shown for comparison (data from this study). DMM composition is after69. The model description is provided in the section Methods. Diagrams showing the Kermadec arc glasses and modeling results of primary Au vs. (A) 1000x(Au/Cu) and B 1000x(Ag/Cu).
It should be noted that the presence of an active back-arc spreading center behind the Kermadec arc should favor the involvement of depleted mantle from beneath the back-arc in repeated hydrous melting events under the volcanic front66. A depleted, hydrous subarc mantle that underwent ca. 5% previous melt extraction has also been proposed to explain the fluid-immobile trace element distribution at Monowai located ~200 km to the north of Putoto (e.g., ref. 67). Multiple mantle depletion has also been invoked to explain high Au and high Au/Cu in boninitic glasses from the Semail ophiolite20, high Pt/Ru (which show similar melting behavior to Au/Cu) in the most depleted melts at the Fonualai spreading center in the Lau Basin19, and generally high Pt/Ru ratios in boninites9.
Taken together, Au contents and their systematics with other chalcophile elements in the Kermadec arc glasses are different from those in MORB and requires specific conditions of magma generation. Almost all the Kermadec arc glasses require extensive high-temperature melting of variably depleted hydrous subarc mantle in the presence of very minor amounts of sulfide liquid or beyond the sulfide stability limit. Furthermore, super-mantle Au/Cu values in about half of the samples requires re-melting of depleted mantle under hydrous fluid-flux from the subducting plate. Our data also suggest that the magmatic Au endowment and its availability for formation of hydrothermal sulfide deposits is regionally controlled by the mantle source and physicochemical conditions of its melting. No significant contribution of Au from the subducting slab is required to explain the most Au-rich Kermadec magmas as previously noticed by several studies20,58,60,61,61. The significance of sulfide recycling in the lower oceanic crust similar to the model proposed to form Au-rich magmas in a continental arc setting23 can readily occur locally and preferably in regions of thickened crust. It cannot, however, explain regional association of Au-rich magmas with the most depleted mantle at the Kermadec Arc and possibly at oceanic arcs globally.
Methods
Modeling
Modeling of Au, Ag, Cu and S behavior during partial mantle melting shown in Figs. 4 and 5 has been performed using incremental melting model presented by ref. 15. The increment step was set to 1 wt% melt in all cases. The partition coefficients between sulfide liquid and silicate melts (DSL/SM) and mono-sulfide solid solutions and silicate melts (DMSS/SM) were assumed to be constant during melting: DSL/SM = 1000, 1000, 10000 and DMSS/SM = 280, 40 and 180 for Cu, Ag and Au, correspondingly13. Bulk partition coefficient of sulfide-free peridotite for these elements was assumed to be 0.048 and constant during melting15. Sulfur solubility in the presence of sulfide phase was set constant during melting and equal to 1000 µg/g at relatively reduced conditions (fO2 ~ QFM) corresponding to melting under mid-ocean ridges and 2000–4000 µg/g (fO2 ~ QFM + 1 log unit) at more oxidizing conditions corresponding to the mantle wedge melting68. Sulfur content in the mantle source was assumed to be 200 µg/g for unmelted depleted MORB mantle (DMM) and 100–400 µg/g for mantle wedge melting that implies a variable sulfur addition with slab derived fluids or melts into the mantle wedge. Cu, Ag and Au contents in DMM were assumed to be 30 ng/g, 0.08 and 0.001 µg/g, respectively69. Subduction related input of these elements into mantle wedge was assumed negligible.
Analytical details
Most of the 67 samples were taken from pillow rim glasses and some from glassy groundmass. Following sample crushing, fresh glass chips were handpicked and washed repeatedly in deionized water using an ultrasonic bath. Individual glass chips were then mounted in EpoThin (Buehler) epoxy-resin and the stubs polished using a sequence of increasingly fine-grained sand paper, diamond pastes and finally 0.05 µm Al2O3 powder.
Major element and S, Cl and F contents were determined via electron micro-probe analyses (EMPA) at GEOMAR. The analyses were performed using a JEOL JXA8200 wave-length dispersive electron microprobe, conditions and reference materials as described by ref. 70. For trace elements, re-polished mounts (to remove the carbon coating required for the EMPA analyses) were analyzed with laser-ablation inductively coupled mass spectrometry (LA-ICPMS) at Institute of Geosciences, Kiel University. The analyses were performed using a quadrupole-based ICP-MS (Agilent 7900) and a ArF 193 nm Excimer LA system (Coherent GeoLas HD). In-situ micro-sampling was done with 160 µm pit size and 20 Hz pulse frequency at 10 J cm−2 fluence. Analyses were performed using a modified large volume ablation cell (LDHCLAC71, ETH Zürich, Switzerland). The generated aerosol was transported with 0.75 L min-1 He and mixed with 0.6 L min-1 Ar prior to introduction into the ICP. The ICP-MS was operated under robust standard conditions at 1500 W and optimized for low oxide and hydroxide formation (typically ThO/Th ≤0.4% and ThOH/Th ≤0.05% monitored during entire analytical sessions at 248 and 249 m/z). Sixty-five major and trace elements were analyzed on one spot. Analysis included 60 s instrumental background measurement (laser-off) and 60 s signal collection during laser ablation. Dwell time was 40 ms for Se, Ag, Re, Ir, Pt, Au, Tl and Bi and 1 ms for all other elements. One analytical cycle lasted 0.649 s; complete analysis comprised data from 180 cycles. Analytical signal on two mass-numbers was collected for Cu (63, 65), Ag (107, 109), Re (184, 185), Ir (191, 193), and Pt (194, 195); one mass-number for all other elements. Initial data reduction was performed using Glitter software package72 for setting time-resolved integration windows and initial non matrix-matched calibration using SRM-NIST612 glass73 and Ca as internal standard. The subsequent data reduction and matrix-matched calibration was done in Microsoft Excel using in-house made tables. On the first step, correction for isobaric interferences has been applied for Se, Ag, Re, Ir, Pt and Au. For elements in focus of this work – Ag and Au – measured intensities were corrected for interferences of 91Zr16O on 107Ag, 93Nb16O on 109Ag and 181Ta16O on 197Au. To measure the oxide production rates on masses of interest we analyzed in every analytical block of 20 points pantellerite glass – in house reference glass PANT from Pantelleria Island. This glass was chosen as the best one for corrections based on many years of experience with minerals, natural glasses and glasses artificially doped with some elements, because this glass is highly evolved and extremely enriched in all incompatible elements including Zr (2700 ppm), Nb (310 ppm), and Ta (18.6 ppm), and also equilibrated at oxygen-reduced conditions and is sulfide saturated, providing strong depletion in strongly siderophile and chalcophile elements, especially heavy PGE and Au. The glass contains 0.13 ppm Ag (obtained by numerous replicate analyses of Ag using mass 107 and natural zircon to estimate Zr/ZrO rate) and negligible (not detected and assumed to be 0 ppm) amounts of Au. The accuracy of the applied correction is confirmed by equal concentrations obtained for different isotopes (e.g., Ag on 107 and 109 m/z) and also by intercepts of calibration lines close to zero. The data corrected for interferences were then corrected for instrumental drift using NBS NIST612 (S, Se, Cd, In, As, Ag, Sb, Re, Pt, Au, Tl, Bi) and BCR-2G (all other elements) glasses. The final matrix-matched calibration was based on a series of 20 whole-rock international reference materials prepared as nano-particulate powder pellets (NPP) for direct in situ-analysis by laser ablation74 and MPI-DING glasses75 using recommended values from the GeoReM data base (http://georem.mpch-mainz.gwdg.de). During the final step, true concentrations were calculated by matching the sum of major element oxides to 100 wt%76,77,78. NPP BHVO-2 and TDB-1 and MPI-DING glasses (KL2-G, GOR128-G) were analyzed as unknowns in the course of this study for checking long-term stability of the calibration and comparison with other studies (Supplementary Data 3).
Typically, each glass sample was analyzed at 4 to 6 spots. Owing to the large beam-diameter of 160 μm, ablation of microlites was inevitable for some samples. When a significant range of compositions was obtained due to variable entrapment of microlites, the data were corrected to obtain pure glass composition using regressions versus K2O measured by LA-ICP-MS, extrapolated to K2O from separate EPMA data. Where LA-ICP-MS data revealed no statistically significant trends and clustered near constant values within analytical uncertainty, the data were averaged and treated as the composition of the bulk ground mass (Supplementary Data 1).
Accuracy of analyses is estimated from calibration lines as average relative deviation of data for reference materials from their recommended values to be 10% relative for Se (calibration range of 0.027–0.351 ppm, number of reference samples N = 8), Au (0.7–6.3 ppb, N = 5) and Ag (24–139 ppb, N = 9) and 5% for Cu (18.6–343 ppm, N = 13). Analytical detection limits for Se, Ag and Au are dependent on instrument sensitivity and also on the amount of interfering oxides and were estimated to be ~0.02 ppm for Se, ∼1 ppb for Ag and ∼0.2 ppb for Au in Kermadec arc samples, which are depleted to moderately enriched in middle REE, Zr, Nb and Ta. The total uncertainty at 95% confidence level was estimated by using the error propagation rule as square root of sum of squared long-term accuracy of standard measurements and squared two relative standard deviation of sample replicate measurements.
Data availibility
The major- and trace element dataset and related quality control data produced for and used in this study are available for download here: https://data.mendeley.com/datasets/ps5j6w5hjb/1.
References
Sillitoe, R. H. The tops and bottoms of porphyry copper deposits. Econ. Geol. 68, 799–815 (1973).
De Ronde, C. E. J. et al. Submarine hydrothermal activity along the mid-Kermadec arc, New Zealand: large-scale effects on venting. Geochem. Geophys. Geosyst. 8, Q07007 (2007).
Fuchs, S., Hannington, M. & Petersen, S. Divining gold in seafloor polymetallic massive sulfide systems. Mineralium Depos. 54, 789–820 (2019).
Richards, J. P. Magmatic to hydrothermal metal fluxes in convergent and collided margins. Ore Geol. Rev. 40, 1–26 (2011).
Timm, C., de Ronde, C. E. J., Leybourne, M., Layton-Matthews, D. & Graham, I. J. Sources of chalco- and siderophile elements in Kermadec arc lavas. Econ. Geol. 107, 1527–1538 (2012).
Mungall, J. E. Roasting the mantle: Slab melting and the genesis of major Au and Au-rich Cu deposits. Geology 30, 915–918 (2002).
Jenner, F. E., O’Neill, H. C. S., Arculus, R. J. & Mavrogenes, J. A. The magnetite crisis in the evolution of arc-related magmas and the initial concentration of Au, Ag and Cu. J. Petrol. 51, 2445–2464 (2010).
Jenner, F. E. et al. Chalcophile element systematics on volcanic glasses from the northwestern Lau Basin. Geochem. Geophys. Geosyst. 13, Q06014 (2012a).
Hamlyn, P. R., Keays, R. R., Cameron, W. E., Crawford, A. J. & Waldron, H. M. Precious metals in magnesian, low-Ti lavas: implication on metallogenesis and sulfide saturation in primary magmas. Geochim. Cosmochim. 49, 1797–1811 (1985).
Keays, R. R. The role of komatiitc and picritic magmatism ans S-saturation in the formation of ore deposits. Lithos 34, 1–18 (1995).
Zajacz, Z., Candela, P. A., Piccoli, P. M., Sanchez-Valle, C. & Wälle, M. Solubility and partitioning behaior of Au, Cu, Ag and reduced S in magmas. Geochim. Cosmochim. Acta 112, 288–304 (2013).
Botcharnikov, R. E. et al. High gold concentrations in sulfide-bearing magma under oxidizing conditions. Nat. Geosci. 4, https://doi.org/10.1038/NGEO1042 (2011).
Li, Y., Audétat, A., Liu, Z. & Wang, F. Chalcophile element partitioning between Cu-rich sulfide phases and silicate melt and implications for the formation of Earth’s continental crust. Geochim. Cosmochim. Acta 302, 61–82 (2021).
Li, Y. & Audedat, A. Effects of temperature, silicate melt composition, and oxygen fugacity on the partitioning of V, Mn, Co, Ni, Cu, Zn, As, Mo, Ag, Sn, Sb, W, Au, Pb, and Bi between sulfide phases and silicate melt. Geochim. Cosmochim. Acta 162, 25–45 (2015).
Lee, C.-T. et al. Copper systematics in arc magmas and implications for crust-mantle differentiation. Science 336, 64–68 (2012).
Ding, S. & Dasgupta, R. The fate of sulfide during decompression melting of peridotite – implications for sulfur inventory of the MORB-source depleted upper mantle. Earth Planet. Sci. Lett. 459, 183–195 (2017).
Arndt, N., Lesher, C. M. & Czamanske, G. K. Mantle-derived magmas and magmatic Ni-Cu-(PGE) deposits. Econ. Geol. 100, 5–24 (2005).
Hamlyn, P. R. & Keays, R. R. Sulfur Saturation and second-stage melts; application to the Bushveld platinum metal deposits. Econ. Geol. 81, 1431–1445 (1986).
Dale, C. W., Macpherson, C. G., Pearson, D. G., Hammond, S. J. & Arculus, R. J. Inter-element fractionation of highly siderophile elements in the Tonga arc due to flux melting of a depleted source. Geochim. Cosmochim. 89, 202–225 (2012).
Belgrano, T. M. et al. Metallogeny of subduction initiation revealed by chalcophile element behavior in the Samail ophiolite. Earth Planet. Sci. Lett. 667, 119486 (2025).
Prokovsky, G. S. et al. Sulfur radical species form gold deposits on Earth. Proc. Natl. Acad. Sci. USA 112, 13484–13489 (2015).
Kepezhinskas, P., Defant, M. J. & Widom, E. Abundance and distribution of PGE and Au in the island-arc mantle: implications for sub-arc metasomatism. Lithos 60, 113–128 (2002).
Holwell, D. A. et al. Mobilisation of deep crustal sulfide melts as a first order control on upper lithospheric metallogeny. Nat. Commun. 13, 573 (2022).
Sun, W., Arculus, R. J., Kaminetsky & Binns, R. A. Release of gold-bearing fluids in convergent margin magmas prompted by magnetite crystallization. Nature 431, 975–977 (2004).
Park, J.-W. et al. Chalcophile element fertility and the formation of porphyry Cu ± Au deposits. Mineralium Depos. 54, 657–670(2019).
Brandl, P. et al. The origin of magmas and metals at the submarine Brothers volcano, SW Pacific. Econ. Geol. https://doi.org/10.5382/econgeo.4973 (2022).
de Ronde, C. E. J. et al. Intra-oceanic subduction-related hydrothermal venting, Kermadec volcanic arc, New Zealand. Earth Planet. Sci. Lett. 193, 359–369 (2001).
de Ronde, C. E. J. et al. Submarine hydrothermal activity and gold-rich mineralization at Brothers volcano, Kermadec arc, New Zealand: Mineralium Deposita https://doi.org/10.1007/s00126-011-0345-8 (2011).
de Ronde, C. E. J. et al. The Anatomy of a buried submarine hydrothermal system, Clark volcano, Kermadec Arc, New Zeland. Econ. Geol. 109, 2261–2292 (2014).
de Ronde, C. E. J., Humphris, S. E., Höfig, T. W. & Reyes, A. G. Critical role of caldera collapse in the formation of seafloor mineralization: the case of Brothers volcano. Geology 47, 762–766 (2019).
Bassett, D. et al. Crustal structure of the Kermadec arc from MANGO seismic refraction profiles. J. Geophys. Res. 121, https://doi.org/10.1002/2016JB013194 (2016).
Hoernle et al. Age and geochemistry of volcanic rocks from the Hikurangi and Manihiki oceanic Plateaus. Geochim. Cosmochim. Acta 74, 7196–7219 (2010).
Timm, C. et al. Subduction of the oceanic Hikurangi Plateau and its impact on the Kermadec arc. Nat. Commun. 5, 4923 (2014).
Wieser, P., Jenner, F., Edmonds, M., Maclennan, J. & Kunz, B. E. Chalcophile elements track the fate of sulfur at Kilauea Volcano, Hawai’i. Geochim. Cosmochim. Acta 282, 245–275 (2021).
Wilkinson, J. J., Triggers for the formation of porphyry ore deposits in magmatic arcs. Nat. Geosci. 917–925 (2013).
Taylor, B. & Martinez, F. Backarc basin basalt systematics. Earth Planet. Sci. Lett. 210, 481–497 (2003).
Webber, A. P., Roberts, S., Taylor, R. N. & Pitcairn, I. K. Golden plumes: substantial gold enrichment of oceanic crust during ridge-plume interaction. Geology 41, 87–90 (2013).
Pitcairn, I. K. Background concentrations of gold in different rock types. Appl. Earth Sci. IMM Trans. Sect. B 120, 31–38 (2018).
Gill, J. et al. Basalt geochemistry and mangle flow during early backarc basin evolution: Havre Trough and Kermacec arc, Southwest Pacific. Geochem. Geophys. Geosyst. 22, e2020GC009339 (2021).
Hirai, Y. et al. Magnesian andesites from Kibblewhite volcano in the Kermadec arc, New Zealand. J. Petrol. 64, 1–23 (2023).
Hauff, F. et al. R7V Sonne Cruise SO255 „Vitiaz“: An integrated major element, trace element and Sr-Nd-Pb-Hf isotope data set of volcanic rocks from the Colville and Kermadec Ridges, the Quaternary Kermadec volcanic front and the Havre Trough backarc basin, Version 1.0. Interdisciplinary Earth Data Alliance (IEDA), https://doi.org/10.26022/IDEA/111723 (2021).
Simon, A. C. et al. Gold partitioning in melt-vapor-brine systems. Geochim. Cosmochim. Acta 69, 3321–3335 (2005).
Frank, M. R., Candela, P. A., Piccoli, P. M. & Glascock, M. D. Gold solubility, speciation, and partitioning as a function of HCl in the brine-silicate melt-metallic gold system at 800 degrees C and 100 MPa. Geochim. Cosmochim. Acta 66, 3719–3732 (2002).
Zajacz, Z. et al. Alkali metals control the release of gold from volatile-rich magmas. Earth Planet. Sci. Lett. 297, 50–56 (2010).
Prokovsky, G. S. & Dubrovinsky, L. S. The S3- ion is stable in geological fluids at elevated temperatures and pressures. Science 331, 1052–1054 (2011).
He, D.-Y. et al. Mantle oxidation by sulfur drives the formation of giant gold deposits in subduction zones. 11 pp, https://doi.org/10.1073/pnas.2404731121 (2024).
Wysoczanski, R. et al. The tectonomagmatic source of ore metals and volatile elements in the southern Kermadec arc. Econ. Geol. 107, 1539–1556 (2012).
Castillo, P. R., Londsdale, P. F., Moran, C. L. & Hawkins, J. W. Geochemistry of mid-Cretaceous Pacific crust being subducted along the Tonga-Kermadec trench: implications for the genesis of arc lavas. Lithos 112, 87–102 (2009).
Turner, S. et al. 238U-230Th disequilibria, magma petrogenesis, and flux rates beneath the depleted Tonga-Kermadec island arc. Geochim. Cosmochim. Acta 61, 4855–4884 (1997).
Reekie, C. D. et al. Sulfide resorption during ascent and degassing of oceanic plateau basalts. Nat. Commun. 10, 82 (2019).
Danyushesky, L. & Plechov, P. Petrolog3: Integrated software for modelling crystallization processes. Geochem. Geophys. Geosyst. 12, Q070201 (2011).
Zhang, Z. & Hirschmann, M. Experimental constraints on mantle sulfide melting. Am. Mineral. 101, 181–192 (2016).
Jugo, P. J. Sulfur content at sulfide saturation in oxidized magmas. Geology 37, 415–418 (2009).
Wallace, P. & Edmonds, M. The sulfur budget in magmas: evidence from melt inclusions, submarine glasses, and volcanic gas emissions. Rev. Mineral. Geochem. 73, 215–246 (2011).
Li, J.-L. et al. Uncovering and quantifying the subduction zone sulfur cycle from the slab perspective. Nat. Commun. 11, 514 (2020).
Taracsak, Z. et al. Sulfur from the subducted slab dominates the sulfur budget of the mantle wedge under volcanic arcs. Earth Planet. Sci. Lett. 602, 117948 (2023).
Wang, X. Primary sulfur isotopes of intraplate basalts and implications for deep S recycling of altered oceanic crust. Geophys. Res. Lett. 51, e2024GL111829 (2024).
McInnes, B. I. A., McBride, J. S., Evans, N. J., Lambert, D. D. & Andrews, A. S. Osmium isotope constraeints on ore metal recycling in subduction zones. Science 286, 512–516 (1999).
Wang, Z. et al. Metasomatized lithospheric mantle for Mesozoic giant gold deposits in the North China craton. Geology 48, 169–173 (2020).
Grondahl, C. & Zajacz, Z. Sulfur and chlorine budgets control the ore fertility of arc magmas. Nat. Commun. 13, 4218 (2022).
Chang, J., Audedat, A. & Pettke, A. The gold content of mafic to felsic potassic magmas. Nat. Commun. 15, 6988 (2024).
Grove, T., Till, C. B. & Krawcynski, M. J. The role of H2O in subduction zone magmatism. Annu. Rev. Earth Planet. Sci. 40, 413–439 (2012).
Syracuse, E. M., van Keken, P. E. & Abers, G. A. The global range of subduction zone thermal models. Phys. Earth Planet. Inter. 183, 73–90 (2010).
Katz, R. F., Spiegelman, M. & Langmuir, C. H. A new parameterization of hydrous mantle melting. Geochem. Geophys. Geosyst. 4, 1073 (2003).
Perrin, A., Goes, S., Prytilak, J., Rondenay, S. & Davies, D. Mantle wedge temperatures and their potential relation to volcanic arc location. Earth Planet. Sci. Lett. 501, 67–77 (2018).
Woodhead, J., Eggins, S. & Gamble, J. High field strength and transition element systematics in island arc and back-arc basin basalts: evidence for multi-phase melt extraction and a depleted mantle wedge. Earth Planet. Sci. Lett. 114, 491–504 (1993).
Timm, C., Graham, I., de Ronde, C. E. J., Leybourne, M. & Woodhead, J. Geochemical evolution of Monowai volcanic center: new insights in the northern Kermadec arc system. Geochem., Geophys. Geosyst. 12, Q0AF01 (2011).
Jugo, P. J., Wilke, M. & Botcharnikov, R. E. Sulfur K-edge XANES analysis of natural and synthetic basaltic glasses: implications for S speciation and S content as function of oxygen fugacity. Geochim. Cosmochim. Acta 74, 5926–5938 (2010).
Salters, V. J. M. & Stracke, A. Composition of depleted mantle. Geochem. Geophys. Geosyst. 5, Q05004 (2004).
Ponomareva, V. et al. A full Holocene tephrochronology for the Kamtchatsky Peninsula region: applications from Kamchatka to North America. Quat. Sci. Rev. 168, 101–122 (2017).
Fricker, M. B. et al. High spatial resolution trace element analysis by LA-ICP-MS using a novel ablation cell for multiple or large samples. Int. J. Mass Spectrom. 307, 39–45 (2011).
Griffin, W. L. GLITTER: data reduction softwarefor laser ablation ICP-MS. In Laser Ablation ICP-MS in the Earth Sciences: Current Practices and Outstanding Issues, 308–311 (2008).
Jochum, K. P. et al. Determination of reference values for NIST SRM 610-617 glasses following ISO guidelines. Geostand. Geoanal. Res. 35, 397–429 (2011).
Garbe-Schönberg, D. & Müller, S. Nano-particulate pressed powder tablets for LA-ICP-MS. J. Anal. At. Spectrom. 29, 990–1000 (2014).
Jochum, K. P. et al. MPI-DING reference glasses for in situ microanalysis: New reference values for element concentrations and isotope ratios. Geochem. Geophys. Geosyst. 7, Q02008 (2006).
Pettke, T., Halter, W. E., Webster, J. D., Aigner-Torres, M. & Heinrich, C. A. Accurate quantification of melt inclusion chemistry by LA-ICPMS: a comparison with EMP and SIMS and advantages and possible limitations of these methods. Lithos 78, 333–361 (2004).
Kimura, J.-I. & Chang, Q. Origin of the suppressed matrix effect for improved analytical performance in determination of major and trace elements in anhydrous silicate samples using 200 nm femtosecond laser ablation sector-field inductively coupled plasma mass spectrometry. J. Anal. At. Spectrom. 27, 1549–1559 (2012).
Timm, C. et al. Major and Trace Elements in Kermadec Arc and MORB Glasses. Mendeley Data, V1 https://doi.org/10.17632/ps5j6w5hjb.1 (2026).
Smith, W. H. F. & Sandwell, D. T. Global sea floor topography from satellite altimetry and ship depth soundings. Science 277, 1956–1962 (1997).
Hayes, G. P. et al. Slab2, a comprehensive subduction zone geometry model. Science 362, 58–61 (2018).
DeMets, C., Gordon, R. G., Argus, D. F. & Stein, S. Effects of recent revisions to the geomagnetic reversal time scale on estimates of current plate motions. Geophys. Res. Lett. 21, 2192–2194 (1994).
Haase, K. M. et al. Formation of island arc dacite magmas by extreme crystal fractionation: an example from Brothers Seamount, Kermadec Island Arc, New Zealand. J. Volcanol. Geotherm. Res. 152, 316–330 (2006).
Jenner, F. E. & O’Neill, H. Analysis of 60 elements in 616 ocean floor basaltic glasses. Geochem. Geophys. Geosyst. 13, Q02005 (2012b).
Acknowledgements
We thank the captains and members of the crew of the R/Vs Tangaroa, Ka’imikai-o-Kanaloa and Sonne for expert handling of the ships and equipment, and the pilots of Pisces V for enabling diverse sampling to be accomplished. We thank Henry Gard and Matthias Witte for their help preparing the samples and Mario Thöner and Ulrike Westernströer for analytical support and expertise. Discussions with Fabio Caratori Tontini, Steve Peircy, Yoshihiko Tamura and Jim Gill helped to shape the ideas. We also thank Colin Devey who supplied MORB samples from the Gakkel Ridge. SO255 samples were collected using grant #03G0255A for the SO255-Vitiaz project funded by the German Federal Ministry of Education and Research (BMBF) and GEOMAR. CdR have been funded from grants made by the New Zealand Ministry of Business, Innovation and Education to GNS Science. This work was funded by the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement #79308 (Project GoldTrace) to CT.
Funding
Open Access funding enabled and organized by Projekt DEAL.
Author information
Authors and Affiliations
Contributions
C.T. and M.P. developed the idea and wrote the ms. M.P. developed the analytical protocol and produced the data together with C.T.; C.dR., M.H., K.H., P.B. and M.L. contributed to the writing and clarification of the proposed model. C.T., M.P., C.dR., K.H. conceived, got funding for and carried out the ship expeditions. D.G.S., D.L.M., N.M.S., R.D.T. and R.A. contributed additional samples, data and input to the ms.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
. Communications Earth & Environment thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: Alireza Bahadori. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Timm, C., Portnyagin, M., de Ronde, C.E.J. et al. Hydrous multi-stage mantle melting controls gold enrichment in mafic Kermadec arc magmas. Commun Earth Environ 7, 281 (2026). https://doi.org/10.1038/s43247-026-03338-w
Received:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1038/s43247-026-03338-w
