Introduction

Submarine volcanism is an important source of metals that has regulated the ocean chemical composition, affecting ore deposition and the availability of nutrients throughout Earth’s history (Baker et al. 1995; Stüeken 2020). Submarine hydrothermal vents are known to have operated throughout geological time since the Palaeoarchaean (from at least 3.5 Ga), in response to periods of enhanced volcanic and tectonic activity (Huston et al. 2010; Mercier-Langevin et al. 2014). Areas of active submarine volcanism and hydrothermalism are the site of deposition of volcanic-hosted massive sulphide (VHMS) ore, an important source of Cu, Zn and Pb worldwide, and can host thriving ecosystems (Galley et al. 2007; Lüders et al. 2001). Similar environments are believed to have hosted some of the earliest ecosystems on Earth (Georgieva et al. 2021; Golding et al. 2011; Hofmann 2011).

Several studies have established a link between the tectonic setting, the rock association and the type of commodities present in VHMS deposits: mafic-dominated volcanic successions emplaced at mid-ocean ridges or in back-arc basins tend to have Cu–(Zn)-rich ore, whereas felsic-dominated successions at arc and back-arc centres have Zn-rich ore, relative to mafic-hosted deposits (Franklin et al. 2005; Galley et al. 2007).

The Hokuroku region of north-eastern Japan is the type locality for “Kuroko” (black ore) type VHMS Zn–Pb–Cu deposits, and has been one of the major mining districts of Japan since its discovery in 1861 (Horikoshi 1990; Ohashi 1920; Tanimura et al. 1983), with a total production estimated at 90 Mt of ore (Ross and Mercier-Langevin 2014). Kuroko deposits, with their distinctive association with tuff and lava domes, are the archetype of bimodal-felsic-associated VHMS worldwide and have attracted substantial scientific attention (Ohmoto 1983; Sato 1977). The Kuroko in this region was deposited in association with back-arc bimodal volcanism that resulted from opening of the Japan Sea in the early to mid-Miocene (Jolivet et al. 1994).

Understanding the relationships between volcanism and hydrothermalism leading to deposition of VHMS requires placing quantitative constraints on the pre-eruptive magma storage and evolution, including the behaviour of trace metals (Zn, Cu and Pb) in magmas. However, the study of volcanic rocks in submarine hydrothermal areas is hampered by the nearly ubiquitous alteration, which masks some of the primary magmatic characteristics (Large 1992). In altered and mineralised samples, melt inclusions, if preserved, can be used to estimate the composition of the silicate melt, which cannot be inferred from analyses of bulk samples and groundmass (Audétat 2015). Melt inclusions form when small droplets of silicate melt are trapped in crystals growing in a magma, either individually or together with other magmatic phases such as minerals, fluids and sulphide melt (Larocque et al. 2000). As crystal growth and trapping of silicate melt can occur continually throughout magma evolution, each melt inclusion is representative of the local and instantaneous status of the magma reservoir (Kamenetsky and Kamenetsky 2010). Thus, melt inclusions can be used to evaluate the evolution of chemical composition through crystal fractionation, mixing or other magmatic processes, the fluid-saturation conditions and the development of mineral phases.

In this study, we used textural observations, whole-rock, mineral and melt inclusion analyses and thermobarometric estimates of volcanic rocks representing the succession hosting the Kuroko Zn–Pb–Cu VHMS to place constraints on the compositional and thermal evolution of the magmas, the depth of magma storage and the influence of these parameters on the eruption style. In addition, we used well preserved, glassy melt inclusions to establish the primary magmatic metal contents, and to infer the conditions leading to the development of submarine back-arc volcanism that hosts the mineralisation.

Geological background

Deposition of the Kuroko ore of Japan is closely related to the tectonic activity associated with the opening of the Japan Sea and accompanying Miocene back-arc volcanism, which is responsible for the deposition of a suite of rocks referred to as the Green Tuff Belt (Horikoshi 1990; Sato 1977). Within this area, the Hokuroku region, a fault-bounded ca. 30 × 40 km basin, is the most richly mineralised and is the type locality of this ore type. These VHMS deposits can be classified based on their host succession as bimodal-felsic type (Barrie and Hannington 1999; Franklin, et al. 2005). The basement of the Japanese archipelago separated from mainland Asia in the early–mid Miocene, with a clockwise rotation of south–west Japan, and anti-clockwise rotation of north–east Japan (Jolivet et al. 1994). Continued extension caused arc-parallel continental rift basins, including the Aosawa, Babame and Hokuroku Rifts (Fig. 1A).

Fig. 1
figure 1

A Schematic map of north–east Japan with the distribution of mid-Miocene continental rift basins and ridges, Kuroko deposits (round symbols) and the position of the volcanic front at around 14 Ma; the boxed area indicates the Hokuroku district. B Simplified stratigraphic log of the Hokuroku district and location of samples (yellow spots). Modified from Morozumi, et al. (2006); Yamada and Yoshida (2013); Yoshida, et al. (2013)

Subsidence of the Hokuroku basin along normal faults occurred during the late stages of extension, and is thought to have reached a maximum water depth below the Miocene calcite compensation depth, with bathyal conditions persisting from well before to well after ore deposition (Guber and Merrill 1983; Ohmoto and Takahashi 1983). The Hokuroku rift was filled with a bimodal, felsic-dominated succession of basalt, andesite, and felsic lava, domes and tuff, as well as mudstone, in which the felsic component represents around 80% of the erupted volume (Yamada and Yoshida 2011). The Kuroko formed during the waning stages of this phase of volcanism. Development of submarine calderas at the time of ore deposition has been proposed, chiefly based on sudden lateral thickness variations of volcanic units (Ohmoto 1996; Ohmoto and Takahashi 1983), similarly to what is observed in modern equivalents (de Ronde et al. 2014). Formation of calderas as the loci of discharge of hydrothermal fluids, and deep marine conditions (Guber and Merrill 1983), suppressing fluid boiling, are considered important factors in the formation of Kuroko ore (Ohmoto and Takahashi 1983). Mine-scale geological observations indicated that Kuroko deposits of the Hokuroku district are associated with shallow felsic intrusions and resurgent domes intruding pyroclastic deposits, and associated breccia deposits (Horikoshi 1969). In cross section, Kuroko ore either overlies a dome, or is situated between two domes (Hashiguchi 1983; Tanimura et al. 1983).

The known Kuroko deposits in the Hokuroku area appear to be limited to a narrow stratigraphic interval, and to have formed over a narrow time span. Re–Os dating of Zn-rich black and Cu-rich yellow ore provided an age of 14.32 ± 0.51 Ma (Terakado 2001). Lead isotope compositions of the ores mostly overlap with the volcanic rocks of the host succession, and only partly with the more isotopically variable basement metasedimentary rocks, indicating Pb derivation mostly from the volcanic rocks (Fehn, et al. 1983; Urabe and Marumo 1991). A wide variation of Fe isotopic values of ferruginous exhalates associated with the ore and the overlying volcanic units has been interpreted as indicating local suboxic bottom water conditions during and after ore deposition (Otake, et al. 2021).

Sampling and analytical methods

Samples representative of volcanic units of the succession in the Hokuroku area were collected from outcrops for textural observations, whole-rock and mineral analyses and melt inclusion study (Fig. 1B). The samples were prepared as polished thin sections and as polished grain mounts for optical microscope observations. Additional observations were made on unpolished grains that had been cracked to expose fluid inclusions. This approach allowed avoiding water and other solvents during polishing, which may dissolve minerals contained in fluid inclusions, and minimised contamination with polishing powder.

For whole-rock major and trace element analyses, the samples were crushed and then milled in an agate mortar at Akita University. The analyses were carried out at Activation Laboratories, Canada, using inductively coupled plasma optical emission spectrometry (ICP-OES) and inductively coupled plasma mass spectrometry (ICP-MS) for major and trace elements and instrumental neutron activation analysis (INAA) for Au, using their Lithoresearch analytical protocol (https://actlabs.com/geochemistry/lithogeochemistry-and-whole-rock-analysis/lithogeochemistry/). Briefly, the ICP-MS method involves flux melting of powdered samples in a crucible to produce glass beads, part of which are then dissolved for analysis, ensuring full digestion of the samples. The analyses are reported in Table 1.

Table 1 Whole-rock chemical analyses of volcanic rocks of the Hokuroku area

For melt inclusion study, the samples were gently crushed with a steel mortar and pestle and sieved to separate various grain size intervals. Phenocrysts of quartz and plagioclase were hand picked from the grain intervals 0.5–0.8 mm and 0.25–0.5 mm, and mounted in epoxy resin. The mounts were ground with sand paper to expose melt inclusions and polished for microscope observation. Selected grains were lifted from the epoxy using a hot needle and mounted in separate mounts, then C-coated and analysed by various techniques (SEM–EDS, EPMA and LA-ICP-MS). Unheated, naturally glassy melt inclusions were used. The melt inclusion analyses are presented in Supplementary Tables 1 and 2.

Major element contents of glass inclusions, plagioclase and oxides were measured by electron probe (EPMA) using a JEOL 8830 Superprobe at Akita University. The analytical conditions were chosen to reduce element mobilisation during beam interaction with plagioclase and glass, and include 15 kV acceleration, 6 nA beam current and 10 μm defocussed beam size for plagioclase and glass, whereas oxides were measured at 15 kV, 20 nA and focussed beam. Alkalis were measured first. The set of standards used includes albite (for Na, Al and Si), apatite (P, Cl), enstatite (Fe and Mg), K-feldspar (K), wollastonite (Ca), pyrophanite (Ti and Mn) and barite (S). A summary of EPM analyses is presented in Table 2.

Table 2 Plagioclase and Fe–Ti-oxide electron probe analyses and estimates of magmatic parameters Supplementary Tables 1 and 2. Melt inclusion analyses (EPMA and LA-ICP-MS)

Major and trace elements of melt inclusions and host plagioclase were measured by laser ablation ICP-MS using an ASI Resonetics 193 nm Excimer laser ablation system and a Thermo Scientific iCAP ICP-MS at the University of Johannesburg. The laser spot size used was 25 µm diameter, with a fluence of 1.4 J/cm2 and an ablation rate of 5 Hz. All analyses consisted of a 15 s blank signal, followed by 30 s of ablation. The primary reference material used was glass NIST 610 (Jochum et al. 2011) for all elements, except Mn, Fe, Zn and Sn, for which glass BHVO-2 g (Jochum et al. 2005) was used, and glass BCR-2 g (Jochum et al. 2005) was ablated as a control standard. The analyses were recalculated so that the major element oxides (SiO2, TiO2, Al2O3, CaO, FeO, MgO, Na2O, K2O) were total 100 wt%. The program LADR (www. norsci.com) was used for data reduction. A comparison with published values of BHVO-2 g and BCR-2 g indicates accuracy better than 10% relative for most elements, except for Mg, Sn, Y, Zr, Yb, Dy, and Th (better than 20%).

Because of the small size of some melt inclusions, which is comparable to the laser ablation spot size, in some analyses, the host mineral was accidentally ablated together with the glass, resulting in mixed analyses. Partial ablation of the host, in addition to direct observation, can be identified based on the anomalous chemical composition, such as high Al and Ca contents and low Fe and Ti contents in plagioclase-hosted inclusions. To obtain the compositions of the glass only, these mixed analyses were deconvoluted by modelling as a two-end-member-mix of the glass (the unknown in the calculation) and the plagioclase, which was measured in spots adjacent to the inclusion. By varying incrementally the proportion x of glass in the mixed analyses, the calculation allowed to bring the Al values within the range measured by EPMA. For comparable methods, see Agangi and Reddy (2016) and Halter et al. (2002). This correction was applied to both major and trace elements.

Raman spectroscopy was used to identify fluid and mineral phases. A Renishaw InVia Raman system installed at Akita University, equipped with a green laser (532 nm), was operated using laser powers of 10% or less. Acquisitions were made at 600 grating, between 0 and approximately 4000 cm−1. Each analysis was composed of 3–5 acquisitions, each lasting for 5–10 s. The instrument was calibrated using a Si metal standard.

The ratio of bubble and melt inclusion volume was estimated based on photomicrographs, by approximating the shape of bubbles and melt inclusions to the nearest geometric shape to calculate volumes and volume ratios. The uncertainty in this type of estimates may be mostly attributed to an incorrect choice of approximating shape. After entrapment, bubbles can form in melt inclusions because cooling induces shrinking of the melt in comparison to the less compressible host crystals, or due to exsolution of a fluid phase in melts trapped close to fluid-saturated conditions (Yang and Scott 2005). The size of shrinkage bubbles will depend on the relative volume change between the host and the trapped melt during cooling thus, relatively constant ratios of bubble/melt inclusion volumes are expected for such bubbles. In addition, cracking of the host and decompression of the inclusion will result in larger bubbles (Skirius et al. 1990). Visibly cracked melt inclusions were avoided during volume measurements and analyses in this study.

Sample description

The complexity of the local geology is reflected by differences in the geological maps produced for the study area, which include the one by the Geological Survey of Japan (Nakajima 1989), and by the Ministry of Economy, Trade and Industry (METI 2004). In this study, the most recent maps and rock unit classification of Yamada and Yoshida (2011; 2013) were adopted. In the map of Yamada and Yoshida (2013), felsic samples Hok06, 07, 09 and 10 occur in volcanic units R2/T2 (rhyolite lava and tuff), overlying the Kuroko ore. Samples Hok04 (lithic tuff) and Hok05 (felsic lava) are representative of units T4 and R3, respectively, both underlying the Kuroko. Intermediate lava sample Hok02 is representative of the stratigraphically lowest unit, andesite A5 (Fig. 1B).

The felsic samples are composed of massive to eutaxitic textured ash-rich, lithic and crystal-rich tuff and massive-textured felsic lava, and contain plagioclase and quartz crystals in strongly altered pale green fine-grained groundmass (Fig. 2A–C). Tuff samples Hok04, 06, 07 and 10 have variable proportions of crystals, ash and lithics. Juvenile components (ash, crystals and in some cases pumice lapilli, Fig. 2A) dominate the composition of the tuff samples overlying the Kuroko.

Fig. 2
figure 2

Textures of volcanic rocks of the Hokuroku area. A (and insets) Pumice and crystal tuff with crystals of plagioclase. The bedding is marked by flattening of pumice clasts. Sample Hok07, overlying Kuroko ore. B Lithic tuff with dark grey matrix rich in pumice fragments. Sample Hok04, underlying Kuroko ore. C Massive, plagioclase–quartz-phyric lava with intensely altered matrix. Sample Hok09, overlying Kuroko ore

Under the microscope, plagioclase crystals are mostly euhedral and fresh (Fig. 3A and B), despite the strong alteration of the groundmass, with the notable exception of lava dome sample Hok05, and show limited internal zoning under cross-polarised transmitted light and in BSE images. The quartz grains have an anhedral habit, locally with deep embayments, which likely represent corrosion gulfs. In sample Hok10, the plagioclase crystals show fractures filled with variably devitrified and altered glass and bubbles, which records crystal shattering during volatile exsolution and vesiculation (Fig. 3C). Clinopyroxene occurs as pseudomorphically replaced crystals in sample Hok09, although fresh clinopyroxene (with Mg-number Mg# = 66, measured by SEM–EDS) was identified as inclusions in plagioclase (samples Hok09 and 10). Accessory minerals in felsic samples include apatite, magnetite and ilmenite, which were found in the groundmass and as inclusions in phenocrysts, and occasionally zircon, whereas sulphides are absent from all the samples.

Fig. 3
figure 3

Microtextures of volcanic rocks of the Hokuroku area. A Plagioclase in pumice and ash-rich matrix; some clasts with either round or elongate, sheared vesicles are indicated by dashed lines. Sample Hok06. B Plagioclase-rich tuff with a matrix of oriented glass–bubble shards. Sample Hok10. Plagioclase crystals have deep embayments containing glass and vapour bubbles. C Cracks in plagioclase filled with glass and bubbles, recording late-stage volatile exsolution and vesiculation. Sample Hok10. D Resorbed, embayed quartz in spherulitic (arrows) groundmass. The embayments contain glass and vapour bubbles. Sample Hok09

The matrix of samples Hok06 and Hok10 is composed of mm-scale pumice fragments with variable textures (Fig. 3A) and ash showing flow foliation that wraps around crystals (Fig. 3B). Vesicles in some pumice fragments have sub-rounded shapes, whereas others have randomly oriented, extremely elongate shapes, which define a marked lineation (tube pumice). The variable textures and the random orientation of vesicles in each fragment in these samples rules out flattening of pumice fragments during compaction, and suggest formation of elongate vesicles by shearing of vesicular, foamy melt during flow, as expected from fragmentation in the volcanic conduit during eruption.

Sample Hok04, representative of the rocks underlying the ore, is a lithic lapilli tuff with angular, cm-scale clasts of pumice and green, fine-grained lithic fragments (composed of intensely altered ash or mudstone) in a fine-grained igneous matrix composed of abundant small pumice fragments up to 2 mm in size and crystals of plagioclase and minor quartz in a dark grey aphanitic mass (Fig. 2B).

In contrast to tuff samples, sample Hok09 has a massive, coherent texture (Fig. 2C), with phenocrysts of plagioclase, quartz and chlorite-replaced short prismatic pyroxene crystals; the groundmass contains abundant micro-spherulites (Fig. 3D). Sample Hok09 is crossed by green (chlorite) veinlets. Sample Hok05, representative of a felsic lava dome underlying Kuroko at the Kannondo ore body, has an oligophyric texture with largely altered phenocrysts of plagioclase in a microcrystalline quartz–feldspar groundmass.

Andesite sample Hok02 is fine grained, massive textured, and contains randomly oriented crystals of plagioclase (< 500 μm in length) and an interstitial matrix of chlorite and carbonate. Accessory minerals include magnetite and apatite.

Melt and fluid inclusion textural description

Abundant, glassy melt inclusions were identified in both plagioclase and quartz crystals in samples Hok04, 06, 09 and 10. The presence of glassy inclusions indicates rapid cooling of the magma during ascent and eruption, as expected during rapid quenching of volcanic products in contact with sea water, and makes these samples ideal for studying melt inclusions without the need for homogenisation in the laboratory.

The melt inclusions can be overall subdivided into two groups based on size, texture and colour. The inclusions of the first group (type 1) are up to ~ 50 μm in size (and occasionally up to 100 μm), subhedral negative crystal shaped, and contain light-coloured (mostly pink) to colourless glass (Fig. 4A and B). These inclusions, occurring in all samples observed, can be found as isolated individuals or as trails aligned along crystal growth planes. Some of these inclusions contain homogeneous glass (Fig. 4A), others contain single, or multiple, small vapour bubbles typically representing ca. 0.5–5 vol.% of the inclusion (Fig. 4B and C). These bubbles can be interpreted a shrinkage bubbles. Locally, vapour-only inclusions occur in the same trail as glass inclusions, suggesting co-trapping of melt and vapour (Fig. 4D). In addition, sample Hok04 also contains characteristic foam-textured inclusions, formed by numerous, μm-scale bubbles in clear glass (Fig. 4E), which indicates growth of crystals in a volatile-saturated, foamy melt. Although development of multiple vapour bubbles has been observed in visibly decrepitated melt inclusions (Skirius et al. 1990), the melt inclusions in our samples lack evidence of cracking, and are interpreted to represent primary textures. Magnetite and apatite are also present in some melt inclusions (Fig. 4F).

Fig. 4
figure 4

Textures of igneous inclusions in volcanic rocks of the Hokuroku area. A Assemblage of homogeneous glass (gl) inclusions hosted in plagioclase (Pl). A long apatite (Ap) needle can also be seen. Sample Hok06. B Subhedral, plagioclase-hosted inclusion of colourless glass and vapour and a trail of smaller inclusions distributed along a growth zone parallel to the crystal rim. Sample Hok09. C Trail of glass (± single or multiple bubbles) melt inclusions trapped along a growth surface (marked by dashed line) in plagioclase (Pl). Sample Hok10. D Glass + vapour and vapour-only inclusions co-trapped in plagioclase. Sample Hok10. E Anhedral, foam-textured inclusion composed of glass with numerous vapour bubbles hosted in plagioclase. Sample Hok04. F Glass–vapour-magnetite inclusion hosted in plagioclase. The inset show reflected light image of boxed area. Sample Hok04

The second group of melt inclusions (type 2, especially abundant in lava sample Hok09, and occasionally found in Hok06) are anhedral, round to elongate, worm-like and up to more than 300 μm in length (Fig. 5A). The two types of inclusions coexist in single grains of both quartz and plagioclase (Fig. 5B). The textures of type 1 and type 2 inclusions suggest a chronology of entrapment: type 1 inclusions in several cases occur along growth planes and were entrapped during crystal growth (Fig. 4B and C), whereas type 2 inclusions, as well as embayments, indicate entrapment following a later stage of crystal resorption (Fig. 5A–D). A majority of type-2 melt inclusions contain homogeneous brown- to honey-coloured glass, but others contain variable proportions of vapour and cloudy, mostly altered glass (Fig. 5C). The vapour-rich inclusions are composed of multiple coalescing bubbles imparting a semi-opaque appearance in transmitted light. In some crystal embayments that have the same phase assemblage, vapour bubbles coexist with a rim of semi-opaque yellow–green, partly devitrified and altered glass (Fig. 5C and D).

Fig. 5
figure 5

Textures of igneous inclusions in volcanic rocks of the Hokuroku area. A Anhedral, worm-like inclusions of brown homogeneous (type 2) glass (gl) hosted in plagioclase (Pl). The inset shows an inclusion photographed at a lower focussing plane. Sample Hok09. B Brown glass and pink glass–vapour (V) inclusions hosted in quartz (Qtz). Sample Hok09. C Complex assemblage of colourless (type 1) glass, brown (type 2) glass and vapour-rich inclusions. Left inset: vapour-rich inclusion with a rim of altered glass (alt. gl). Right inset: homogeneous colourless glass. Note the quartz overgrowth (marked by arrows) sealing the vapour inclusion in the centre. Sample Hok09. D Embayment containing altered glass and vapour with the same assemblage as inclusion in C. Sample Hok09

Estimates of bubble/melt inclusion ratios in sample Hok09 are shown in Fig. 6. Type 1 inclusions have ratios up to 5.5 vol% (average 2.3 ± 1.4%), with a notable peak at around 2–3 vol% in the histogram. The small and relatively constant values of bubble proportions suggest post-entrapment formation of the bubbles (shrinkage bubbles). Type 2 inclusions have mostly bubbles representing less than 1 vol% of the inclusions (or they lack a bubble entirely), although several of them have large vapour bubbles representing variable proportions of the melt inclusions (up to 63 vol%). This strong variability of vapour/melt ratios indicates heterogeneous trapping of co-existing melt and vapour, and implies that the magma was fluid saturated at the moment of formation of type 2 inclusions.

Fig. 6
figure 6

Ratio of bubble and melt inclusion volumes in volcanic rocks (sample Hok09, plagioclase and quartz hosts). Type 1 melt inclusions are subhedral; they occur in many cases aligned along crystal growth planes and contain colourless pink glass; type 2 are anhedral, tube-like or round, and represent heterogeneous entrapment of glass and vapour in dissolution channels

Opened fluid inclusions hosted in plagioclase and zircon contain fine precipitates of K–Cl (likely sylvite) and Cl–S–K–Na, identified by SEM–EDS. Melt inclusion bubbles in sample Hok10 contain fine mineral precipitates identified as carbonate by Raman spectra (main peak at ca. 1084 cm−1 and minor at ca. 711 cm−1). Sulphate (possibly gypsum) was locally identified in bubbles of melt inclusions with brown glass (strong Raman peak at 1010 cm−1, and minor peak at 661 cm−1). Raman spectra of some of the vapour inclusions have a broad peak corresponding to liquid water (between approximately 3100–3700 cm−1, likely indicating small amounts of condensed liquid water not clearly visible under the microscope), as well as anatase (peaks at 144 (vs), 636 (s) and 514 (w) cm−1) and SO2 (1154 cm−1), in addition to unidentified peaks at 804 and 1197 cm−1.

Chemical analyses of whole-rocks, minerals and glass inclusions

Whole-rock analyses

Plots of selected major element oxides of samples analysed for this study and literature analyses from the Hokuroku district (Yamada and Yoshida 2011; Yamada et al. 2012) and the time-correlative units in the Oga peninsula and back-arc-related early–mid-Miocene samples of the north–east Japan area at large (Sato et al. 2007; Shuto et al. 2006; 2015; Takanashi et al. 2011) are shown in Fig. 7.

Fig. 7
figure 7

Whole-rock and melt inclusion major element compositions of the Hokuroku area, compared with Miocene back-arc-related samples from north–east Japan. A Total alkali vs SiO2; B (K2O + Na2O – CaO) vs SiO2; C FeO/(FeO + MgO) vs SiO2; D CaO vs SiO2. Low-K2O melt inclusion analyses are not plotted for clarity

In the total alkali vs SiO2 plot, whole-rock analyses of new samples having SiO2 = 54.8–80.1 wt% (recalculated on an anhydrous basis) overlap with available analyses from the Hokuroku region, which span the compositional range from basalt to rhyolite (SiO2 = 52.4–82.4 wt%), and define a subalkaline trend with bimodal mafic-intermediate and felsic SiO2 distribution (Fig. 7A). In the modified alkali-lime index (MALI = Na2O + K2O–CaO, wt%) vs SiO2 plot, new and literature analyses show significant scatter, but the felsic samples mostly plot in the calcic and calcic–alkalic fields (Fig. 7B). The FeO/(FeO + MgO) vs SiO2 plot (where FeO denotes total FeO) indicates a moderate Fe-enrichment with increasing SiO2 for mafic-intermediate samples from the Hokuroku district as well as the north–east Japan in general that is mostly confined to the magnesian field (Fig. 7C). The values of alumina saturation index (defined as ASI = Al2O3/(Na2O + K2O + CaO), mol.) increase with increasing SiO2, and span a wide range of values (0.5–1.5). Overall, much of the scatter observed in major element plots is indicative of sample alteration.

Plots of selected trace element compositions are presented in Fig. 8. In the Th/Yb vs Nb/Yb plot, the samples describe a curved trend extending from near-primitive mantle values to values typical of continental arcs (Fig. 8A). Ratios of fluid-mobile trace elements (Ba, Th, Pb) and fluid-immobile elements (Nb, La, Ce, Yb) can be used to track the influence of subduction-derived fluids on the composition of melts produced by mantle melting and their more evolved, intermediate-felsic counterparts. Deviation from primitive mantle values for Hokuroku samples, reflecting enrichment from slab-derived fluids, can be observed in the Ce/Pb vs SiO2 plot, where mafic to felsic samples have Ce/Pb values (up to 6) below primitive mantle values (Fig. 8B). However, mafic-intermediate samples from the wider north–east Japan back-arc region have variable Ce/Pb, reflecting variable degrees of depletion (Ce/Pb > primitive mantle values of 10) and enrichment by slab fluids (Ce/Pb < 10). Enrichment in LILE compared to primitive mantle values can be also seen in a Ba/La vs Th/Yb plot (Fig. 8C). The Pb/La vs La/Yb plot demonstrates that samples of the Hokuroku district have Pb/La values ranging from near-primitive mantle to higher than continental crust, but only moderate La/Yb (5–12), which plot between primitive mantle and continental crust values (Fig. 8D).

Fig. 8
figure 8

Whole-rock and melt inclusion trace element compositions. A Th/Yb vs Nb/Yb (fields after (Pearce 1996); B Ce/Pb vs SiO2; C Ba/La vs Th/Yb; D Pb/La vs La/Yb. PM: primitive mantle, CC: continental crust, MORB: mid-ocean ridge basalt, OIB: ocean island basalt (Sun and McDonough 1989)

Primitive mantle-normalised plots of new felsic samples have enriched large-ion lithophile elements, with Ba and Th values around 100 times the primitive mantle compositions (Fig. 9A). In the same plots, high field-strength elements Nb and Ti have negative anomalies compared to La and heavy REE, respectively. Intermediate sample Hok02 has an overall similar primitive mantle-normalised pattern as the felsic samples with similar negative Nb anomaly, but has more moderate enrichment in LILE. All samples have positive Pb anomalies.

Fig. 9
figure 9

Primitive mantle-normalised (Sun and McDonough 1989) trace element compositions of whole-rocks (A) and melt inclusions (B)

Glass inclusion compositions

Analysis by EPMA of plagioclase- and quartz-hosted melt inclusions indicates high silica rhyolitic compositions spanning a narrow range of SiO2 (77–80 wt%), with CaO = 0.7–1.7 wt%, Na2O = 2.7–4.0 wt%, FeO = 0.8–1.6 wt%, recalculated on an anhydrous basis (Fig. 7). Chlorine ranges from 0.2 to 0.4 wt% in a majority of analyses, although some analyses are Cl-depleted with values down to 0.02 wt%, and S is up to 0.02 wt%. Major element compositions of all the samples largely overlap, and no significant compositional difference was found between quartz-hosted and plagioclase-hosted melt inclusions or between colourless pink (type 1) and brown-honey-coloured (type 2) glass. When plotted together with whole-rock analyses of the Hokuroku area, major element compositions of melt inclusions overlap with the most felsic whole-rock samples, although whole-rock analyses have more scattered alkalis and CaO, likely due to rock alteration. The melt inclusions are metaluminous to slightly peraluminous, with ASI = 0.9–1.1. In the MALI vs SiO2 plot, the melt inclusions have relatively low MALI values in comparison with the bulk-rock analyses, and plot in the calcic field (Fig. 7B).

A subset of melt inclusions in samples Hok04, 06 and 10 have distinctively low K2O (≤ 2 wt%) and mostly low Cl compared to other melt inclusions, in addition to having more scattered FeO, Na2O and SiO2. Anomalously low-K2O rhyolitic groundmass glass has been reported from Soufrière and Mount St Helens, where it was interpreted as a finely crystallised feldspar–quartz assemblage (Blundy and Cashman 2001), although this observation appears inapplicable to our samples. The low-K2O inclusions in our samples show glassy homogeneous textures and appear as slightly darker than other glass inclusions in BSE images (as a consequence of lower K and Fe). Based on the degassed, low-Cl and -Li compositions and the anomalous content in some major elements, we propose that these low-K2O inclusions may have sampled parcels of melt that had exsolved a Cl-bearing (hydrosaline) fluid.

In primitive mantle-normalised plots, trace element compositions of melt inclusions analysed by LA-ICP-MS largely overlap with whole-rock compositions (Fig. 9A and B). Rare earth elements show moderate fractionation (La/YbN = 1–11 for most analyses), and large ion-lithophile elements (Ba, Rb, Th) are enriched with values clustering around 100 times the primitive mantle values. Lead has positive anomaly compared to light REE, although a few analyses have moderate negative anomalies, and Sr has prominent negative anomalies compared to mid-REE (Fig. 9). The Ce/Pb values (mostly 2–6) are lower than primitive mantle (10) and depleted mantle (25), and typical of arc and back-arc volcanic rocks (Fig. 8B), with a few exceptions of Pb-depleted analyses.

The effect of crystallisation on melt composition can be estimated in plots of incompatible vs incompatible elements (e.g. Th vs Y) and ratios of compatible/incompatible vs incompatible elements (Sr/Rb vs Y). In the Th vs Y plot, the melt inclusions from different units describe separate linear trends with positive slopes (Fig. 10A). In the Sr/Rb vs Y plot, the melt inclusion compositions from different units define separate curved trends with decreasing Sr/Rb for increasing Y, which can be modelled by crystallisation of variable amounts of plagioclase and quartz crystallising in cotectic proportions (0.55 plagioclase and 0.45 quartz; Fig. 10B). The plagioclase-melt distribution coefficients used in these calculations are derived from analyses of melt inclusions and host plagioclase, and are in agreement with values available in the literature (Ewart and Griffin 1994); for quartz, incompatible trace element behaviour was assumed. Such modelling for samples Hok06 and 10, for which a larger dataset is available, indicates evolution of melt batches by > 50% fractional crystallisation.

Fig. 10
figure 10

Melt inclusion trace element and volatile compositions and modelled behaviour during magma evolution. A Th vs Y, B Sr/Rb vs Y, C Li vs Y; D Cl vs Y; E Zn vs Y; F Pb vs Y; G Pb vs Zn; H Sn vs Y (Y is considered as immobile during degassing). Dashed lines represent modelled fractional crystallisation (FC). Tick marks indicate 10% intervals of crystallisation. Distribution coefficients for plagioclase were calculated from average plagioclase and glass inclusion analyses (Kd = 5.8 for Sr, 0.03 for Rb, 0.01 for Y, 0.20 for Zn, 0.25 for Pb and 0.003 for Th, except where indicated on the plots)

Lithium ranges from < 0.5 to 72 ppm, and has a broad positive correlation with K2O (this further demonstrates the degassed nature of the low-K2O melt inclusions). In a Li vs Y plot, some analyses show positive correlation (Fig. 10C), reflecting largely incompatible behaviour in the main minerals (feldspar, quartz, magnetite ± pyroxene), although others have low and scattered Li contents, which can be ascribed to Li-loss through degassing. Modelling of Li contents in these latter analyses requires much higher Kd values (up to 2) than inferred from the analysed plagioclase-melt distribution. Some analyses have distinctively low Li < 1 ppm and variable Y and K2O, and plot off the modelled fractionation trends. Considering the high diffusivity of Li in melt and crystals (Charlier et al. 2012), this may be due to post-entrapment diffusive loss of Li from melt inclusions, rather than melt-fluid fractionation. Similarly, Cl concentrations deviate from the trend expected for ideal incompatible behaviour in a Cl vs Y plot, compatible with Cl-loss to a fluid (Fig. 10D).

The behaviour of metals during magma evolution can be evaluated in plots with incompatible elements, such as Y (Fig. 10) and compared with the degassing trends of volatiles (Li and Cl). For example, Zn and Pb display positive correlations with Y, suggesting broadly incompatible behaviour of these elements (Fig. 10E and F). Zinc and Pb have similar mild incompatibility with plagioclase (Kd = 0.2 ± 0.1; Ewart and Griffin 1994), so that their ratio Zn/Pb is expected to be little affected by fractional crystallisation. However, in some samples, Pb deviates towards lower concentrations than expected from fractional crystallisation, which may be attributed to fluid loss. This results in an increasing Zn/Pb in the melt with progressing crystallisation (increasing Y; Fig. 10G). In contrast, Sn, Cu and Mo have more scattered and broadly decreasing values when plotted against Y, which suggests substantial loss during degassing (Fig. 10H). The metal contents of whole-rock samples and melt inclusions are compared in Fig. 11.

Fig. 11
figure 11

Upper continental crust-normalised (Rudnick and Gao 2014) metal contents of whole-rocks and melt inclusions of the Hokuroku district

Mineral chemical analyses (plagioclase)

EPMA analyses indicate that plagioclase phenocrysts are mostly andesine (anorthite content mostly An32-50 in samples Hok06 and Hok09, An34-46 in sample Hok10 and An34-41 in sample Hok04) with limited intragranular zoning. However, irregular zones identified in a few crystals of sample Hok06 reach anorthite contents of An57-60 (Fig. 12).

Fig. 12
figure 12

Plagioclase compositions (An, EPMA) and calculated plagioclase compositions in equilibrium with melt inclusions (indicated by shaded areas, calculated according to Scruggs and Putirka 2018). Note population of plagioclase analyses in clear disequilibrium with the melt in sample Hok06, indicated by striped pattern

Calculation of the composition of plagioclase in equilibrium with melt inclusion compositions (using the method of Scruggs and Putirka (2018)) indicates plagioclase compositions of An38–43 for samples Hok06, An35–43 for Hok09, An40–42 for Hok04 and An36–38 for Hok10. These compositional ranges, although narrower than the analysed plagioclase compositions, overlap with the peaks of distribution of measured plagioclase crystals, indicating that the phenocrysts were in equilibrium with the carrying liquid, with the exceptions of a few analyses from sample Hok06. This substantial equilibrium is also confirmed by the absence of disequilibrium textures, such as sieve textures and reverse zoning of plagioclase, and allows for the application of plagioclase-melt thermometers and hygrometers (Cashman and Blundy 2013).

Estimates of crystallisation temperature, fO2 and water content of the melt

Temperatures were calculated based on plagioclase-melt equilibrium, using equation 24a of Putirka (2008), excluding compositions higher than An52 in sample Hok06, which are not in equilibrium with the melt. The estimates range from 898 to 919 °C for sample Hok09, 922 to 939 °C for Hok06, 884 to 897 °C for Hok10 and 928 to 936 °C for Hok04 (Table 2).

Water estimates were obtained based on two approaches: first, the equilibrium between plagioclase compositions and the average melt inclusion composition of each sample was considered (using the equation 25b of Putirka 2008; Fig. 13A). The average melt inclusion compositions were chosen given the limited major element compositional variation of melt inclusions, and because several plagioclase crystals analysed do not have melt inclusions. Limited plagioclase analyses that show compositional evidence for disequilibrium with the melt (sample Hok06 in particular) were omitted from these calculations. The estimates indicate largely overlapping values between samples of pyroclastic rocks (Hok04, 06 and 10), with H2O = 1.5–2.6 wt%. Water estimates for lava sample Hok09 also overlap with pyroclastic samples, but extend to lower values (≥ 1.3 wt%); the histogram of this sample has a skewed distribution towards lower values, and a broad peak at around 1.5 wt% H2O. In addition, a similar procedure was applied to plagioclase-hosted melt inclusions and their hosts analysed next to the inclusions, resulting in H2O and temperature estimates for single melt inclusion–host pairs. These estimates of H2O range from 1.0 to 3.0 wt%, a somewhat larger range than the previous approach, and temperatures of 868–939 °C. The plot of H2O vs temperature estimates for single melt inclusion–host pairs shows scatter for most samples, with the exception of lava sample Hok09, which has decreasing H2O values for decreasing temperature (from 923 to 868 °C), suggesting degassing during progressive cooling (Fig. 13B).

Fig. 13
figure 13

A Estimates of water content obtained from plagioclase-melt equilibrium using equation 25b of Putirka (2008); average and standard deviation are also indicated for each sample as circle symbols; B plot of H2O vs temperature estimates obtained from melt inclusion–host pairs using equations 25b and 24a of Putirka (2008), respectively. Arrow indicates the cooling-depressurisation trend in lava sample Hok09

To estimate the fO2 conditions of the magma, the magnetite–ilmenite geothermometer and oxy-barometer of Andersen and Lindsley (1988) were applied to magnetite–ilmenite pairs co-trapped in phenocrysts analysed by EPMA. The estimated temperatures, based on 12 oxide pairs in samples Hok06, 09 and 10, are 872–896 °C and fO2 − 10.9–− 11.9, corresponding approximately to FMQ (fayalite–magnetite–quartz buffer) + 1.5 log units.

The composition of cotectic melts in the granitic system (Ab–Qtz–Or) is strongly dependent on the equilibration pressure, and can be used to estimate the depth of magma storage for feldspar–quartz-bearing magmas, if water saturation can be assumed (Gualda and Ghiorso 2013). The CIPW mineral norm compositions of melt inclusions have been calculated using the software Norm4 by Kurt Hollocher (https://minerva.union.edu/hollochk/c_petrology/other_files/norm_calculation.pdf) and plotted in the albite–quartz–orthoclase granitic triplot (Fig. 14). Using melt inclusions compositions, instead of whole-rock analyses, allows circumventing issues due to: 1) element remobilisation during alteration, which is ubiquitous in the Hokuroku area, and 2) crystal accumulation, which can cause the whole-rock compositions to depart from true melt compositions. To account for the effect of the normative anorthite component, mostly ranging from 1 to 6 wt%, the correction of Blundy and Cashman (2001) has been applied to our analyses. The corrected normative compositions of glass inclusions plot around the thermal minimum at 0.5 kbar in the Ab–Qtz–Or triplot (Fig. 14), and describe a curved trend starting at relatively unevolved compositions (norm orthoclase ranging ~ 10 – 20% on a Ab + Qtz + Or = 100 basis, for samples Hok04, 06 and 10), and extending towards more evolved compositions (norm orthoclase up to > 20%, sample Hok09). The first part of this trend suggests near-isobaric fractionation at low pressures (< 1 kbar), corresponding to shallow crustal depths of approximately 4 km or less. For more evolved compositions (norm orthoclase > 20%), the trend deviates towards even higher normative quartz, suggesting decreasing equilibration pressures.

Fig. 14
figure 14

Quartz–albite–orthoclase ternary plot of normative melt inclusion compositions. The analyses have been corrected for their An normative content according to the method of Blundy and Cashman (2001). The grey lines indicate the phase boundaries at variable H2O pressures (indicated for each line). The thick black line indicates the isothermal decompression path calculated by Blundy and Cashman (2001); precipitating phases (Fls—feldspar, Qtz—quartz) and pressures at various points on the trajectory are indicated

Discussion

Evidence for fluid saturation of the melt and multiple exsolution events

The samples studied here show ample textural evidence of magma degassing, vesiculation and fragmentation. This evidence includes: 1) sheared and unsheared pumice fragments (lapilli and ash) observed in tuff samples (Fig. 3); 2) melt inclusions and embayments with co-trapped silicate melt and vapour (vapour-dominated inclusions, and micro-vesicular foam texture, Figs. 4, 5); and 3) glass–vapour-filled microcracks in phenocrysts (Fig. 3C).

In addition to textures, chemical evidence for fluid exsolution is provided by variable Cl and Li contents of melt inclusions, and by the Cl–Y and Li–Y relationships, which indicate volatile loss through degassing. Experiments on water-saturated granitic melts predict that a melt having the measured Cl and estimated H2O contents found in our samples would have been saturated with a hydrosaline fluid, a vapour, or both at low pressure down to approximately 0.5 kbar (Webster 2004) (Fig. 15). The trends of Cl and Li with incompatible elements (Fig. 10C and D), and the fine-grained K–Cl ± Na ± S precipitates observed in fluid inclusions provide direct evidence for such a saline fluid. Exsolution of a hydrosaline fluid has the potential to affect the behaviour of metals during degassing (Kouzmanov and Pokrovski 2012).

Fig. 15
figure 15

Melt inclusion Cl contents measured by EPMA vs H2O estimates. Experimental fluid saturation line from Webster (2004)

Tuff-forming explosive eruptions are believed to be initiated by vapour exsolution from the melt (Sparks et al. 1994). However, in deep submarine environments, the overlying water column imposes a confining pressure that is expected to limit the ability of magma to vesiculate and expand (Busby 2005), and the mechanisms leading to pyroclastic eruptions in such settings is still debated (Carey et al. 2018).

The microtextures observed in both tuff and lava samples indicate vapour exsolution both before and during eruption and record the early stages of the process that would have led to production and expansion of pumice during eruption (Figs. 4, 5). Thus, the fundamental triggers for tuff-forming and effusive eruptions appear to be similar. The major element compositions of melt inclusions in tuff and lava samples are undistinguishable, and the plagioclase-melt temperature estimates overlap, indicating that the contrasting eruptive behaviour was not controlled by the chemical composition or the temperature of the melt. However, lower average water estimates in rhyolitic lava sample Hok09 (Fig. 13) compared to tuff samples suggest an important role for dissolved H2O on eruptive behaviour (Roggensack et al. 1997). Water content estimates in sample Hok09 (average 1.5 ± 0.2 wt%) have a skewed distribution towards the lower end (with a broad peak between 1.3 and 1.7 wt%, Fig. 13A). This is compatible with the decreasing H2O-temperature trend (Fig. 13B) and the decompression trend estimated from the haplogranite plot (Fig. 14). In addition, the contrasting textures of type 1 and type 2 melt inclusions in sample Hok09 also suggest progressive degassing. Type 1 melt inclusions contain homogeneous glass or small bubbles (< 5 vol.%), which is compatible with rapid quenching and bubble formation during cooling (shrinkage bubbles). In contrast, late-forming, type 2 inclusions contain variable, and locally large, vapour fractions, indicating melt vapour co-trapping during extensive degassing (Fig. 6).

In felsic magmas, whose high viscosity may cause retention of bubbles (closed-system degassing), degassing and vesiculation can be regarded as a ‘runaway’ process that leads to expansion and eventually explosive eruptions (Cassidy, et al. 2018). If such closed-system conditions are retained, once volatile saturation is reached (which is mostly controlled by pressure), expansion of gas causes the foamy magma to rise, thus causing further decompression and degassing. Eventually, in this self-propagating and accelerating process, the increasing rate of expansion will exceed the limits of brittle behaviour and cause breakage, i.e. pumice fragmentation and explosive eruption (Dingwell 1996), as is typical in Neptunian eruptions (Allen and McPhie 2009).

However, the microtextures observed in our Hokuroku samples indicate that fluid saturation was reached at different stages of magma evolution. After formation, some vapour bubbles were trapped by overgrowing host quartz and plagioclase (Figs. 4, 5C) implying that enough time passed between vesiculation and eruption for further crystal growth, whereas other bubbles are present in open embayments connected with the groundmass (Fig. 5D) and in cracks (Fig. 3C), indicating the presence of this fluid immediately prior to and even during eruption. Taken together, this evidence is suggestive of multiple exsolution events in single samples, the last of which led to eruption. Thus, exsolution accompanied crystallisation, and fluid-saturated, bubbly magma existed for an undetermined amount of time without leading to immediate eruptions.

Shallow magma storage and evolution at thermal-minimum (cotectic) conditions

Coexisting feldspar and quartz, and the normative compositions of melt inclusions in this study suggest crystallisation at, or near, the granitic thermal minimum at low water-saturated pressure (PH2O < 1 kbar, and locally < 0.5 kbar; Fig. 14), and the spread of their trace element compositions indicates extensive fractionation (Fig. 10A and B). The presence of pyroxene (this study; Yamada and Yoshida 2011) instead of amphibole is also indicative of crystallisation pressure lower than approximately 1 kbar (depending on temperature; Blundy and Cashman 2001) and the low water estimates provide further evidence of low-pressure equilibration (Moore et al. 1995). Such shallow, near-cotectic magma storage and crystallisation conditions are observed for all the units studied here, which range in age from pre- to post-ore deposition.

In the Hokuroku samples, the homogeneous major element melt compositions and the absence of disequilibrium textures of plagioclase indicate equilibrium between crystals and melt, and suggest a compositionally homogeneous magma reservoir. The limited compositional ranges of plagioclase phenocrysts (with a few exceptions) appear to rule out major events of mixing as a mechanism of magma evolution and as a trigger for eruption, as proposed instead for several arc volcanos (Pallister et al. 1992; Sparks et al. 1977). Even the resorption textures observed in some phenocrysts (forming embayments and anhedral, tubular type 2 melt inclusions; Fig. 5) were not due to injection of more mafic and hotter magma, as indicated by the glass compositions in these inclusions, which is undistinguishable from other glass inclusions. Instead, these re-melting events may have been induced by a shift of the thermal minimum towards the quartz apex during decompression.

Such textural and compositional homogeneity contrasts sharply with observations from typical arc volcanoes, which contain frequent disequilibrium textures, commonly carrying compositionally and texturally dissimilar crystals that do not share the same growth history (Cashman and Blundy 2013; Davidson et al. 2007), as well as mafic and intermediate enclaves and banded pumices indicating mingling during injection of hot mafic magma into a cooler, more felsic resident magma, a process that is also credited for triggering re-activation of crystal mushes and eruption (e.g. at Aira Caldera; Geshi et al. 2020). These complex assemblages found in arc lavas indicate mixing of magmas formed over wide ranges of depths and temperatures, providing evidence for complex trans-crustal plumbing systems (e.g. Bishop Tuff; Anderson et al. 2000; Santorini; Cadoux et al. 2014). The apparent lack of evidence for deep roots in the Hokuroku sub-volcanic plumbing system may indicate rapid rise of felsic magmas towards the shallow crust, without stalling in the middle crust, perhaps favoured by the overall extensional tectonic setting.

Tectonic influence on the development of shallowly emplaced felsic magma chambers

The depth of emplacement of a magma is controlled by a complex interplay between the rheology of the crust and magma characteristics, such as buoyancy and water content, and the most favourable conditions for magma stalling in the shallow crust have been estimated around 2 ± 0.5 kbar (Huber et al. 2019). According to this model, at the depth corresponding to such pressure, the balance between magma addition from below and magma withdrawal through vesiculation and eruption allows growth of large reservoirs, whereas, at shallower depths, extensive vesiculation and eruption prevent magma accumulation. An important implication is that the initial H2O content and the corresponding pressure and depth of fluid saturation of magmas would affect their depth of emplacement. Indeed, a recent comparison of water measurements and estimated depths of magma emplacement of 62 arc volcanos based on geophysical methods suggests that arc magmas stall in the crust at the depths of volatile saturation, and not the depths of natural buoyancy (Rasmussen et al. 2022). Such behaviour may derive from a viscosity increase caused by degassing, implying that the initial H2O content is a key control on the depth of magma emplacement.

In comparison with convergent settings (volcanic arcs), which are characterised by comparatively high H2O contents (5–6 wt% in felsic magmas; Wallace and Bergantz 2005), magmas in extensional settings typically have relatively low H2O contents. For example, analyses of melt inclusions from large-volume high-silica (> 75 wt%) intracontinental rhyolites of the Yellowstone Caldera have yielded H2O of 1.0–2.5 wt% (Befus et al. 2012). Such low-H2O magmas can rise to shallow levels in the upper crust before reaching volatile saturation and vesiculation, which eventually leads to withdrawal through eruption. This can produce large volumes of shallowly emplaced felsic magmas that fractionate towards high-silica compositions. Similarly, the low equilibration pressure estimates for the Hokuroku samples studied here indicate stalling of magma in the very shallow crust, implying that the magma did not saturate at greater depth. This may be the consequence of a low initial water content of the magma, which prevented exsolution and magma withdrawal at deeper crustal levels.

Link between magmatic activity and ore deposition: pre-eruptive metal contents and behaviour during magma evolution and degassing

The influence of metal abundance of magmas that give rise to VHMS deposits is debated. Hydrothermal fluids that form VHMS deposits may derive their metal contents either from interaction with, and leaching of rocks that underlie the deposit, or from exsolution of magmatic volatiles from shallowly emplaced magma (Marques, et al. 2020; Ohmoto 1983; Urabe and Marumo 1991; de Ronde, et al. 2011 and 2014; Agangi, et al. 2018), or a combination of the two. In the next paragraphs, we evaluate the pre-eruption metal endowment of magmas forming the volcanic succession that hosts the Miocene Kuroko VHMS of north–east Japan on the basis of melt inclusion analyses, and estimate the metal partitioning between silicate melt and coexisting fluid.

Overall, our analyses indicate that the magmas had metal contents comparable to average upper continental crust for Zn, and rather depleted Cu, Pb, Sn and Co, whereas Ag and Mo were locally enriched (Fig. 11).

Melt inclusion analyses indicate that metals were lost to an exsolving fluid during prolonged magma degassing in the shallow crust. These metal-bearing fluids would have then risen towards the surface and mixed with hydrothermal circulation cells in the overlying crust. This scenario is mirrored by models of some modern submarine systems (de Ronde et al. 2011; Yang and Scott 2005).

However, our analyses indicate contrasting degassing behaviour of different metals. In particular, melt inclusion analyses show that Zn was retained in the melt during magma evolution, i.e. it had incompatible behaviour (Fig. 10E). Lead shows more compatible behaviour compared to the expected fractionation trend (Fig. 10F), which, in the absence of sulphide in any of the samples, is suggestive of scavenging by a magmatic fluid phase. Other metals, such as Sn, Mo and Cu, have more compatible behaviour, and appear to have been released to exsolving fluids to a significant extent. Such contrasting behaviour opens up the possibility that different metals were contributed to the mineralising fluid by different processes. Retention of metals in the melt (particularly Zn) would have made the resulting volcanic rocks a good source for leaching by circulating hydrothermal fluids, whereas degassing of other metals (such as Cu and, to some extent, Pb) would make magmatic fluids good sources for such elements. This contrasting metal behaviour helps explain the fractionation between Zn and Cu in Kuroko deposits, which results in metal zoning, with the formation of Zn-rich black ore and Cu-rich yellow ore (Sato 1977). Differences in metal contents have been attributed to a difference of temperature (Ohmoto 1996), although this model cannot account for the different Pb-isotope compositions between black and yellow ores. The high-temperature, Cu-rich samples (yellow ore) have unradiogenic Pb compositions, which fully overlap with magmatic rocks, whereas the lower-temperature, Zn-rich black ore has Pb-isotopes that extend to slightly higher values, allowing for a metal contribution from leaching of the more radiogenic Cretaceous basement (Fehn, et al. 1983; Urabe and Marumo 1991). The combination of these observations allows formulating a modified genetic model for Kuroko deposits, which involves formation of black ore by leaching of metals by circulating sea-water-derived fluids, followed by formation of the higher-temperature yellow ore during stages of enhanced magmatic degassing (Fig. 16).

Fig. 16
figure 16

Modified model of Kuroko formation. Shallowly emplaced felsic magma crystallises and degasses, providing metal-bearing fluids that can be injected into overlying hydrothermal circulation cells. Normal faults may have provided a path for both magmas and fluids (modified from Sato 1977)

Finally, an important control on hydrothermal activity is represented by the shallow level of magma emplacement. Storage, crystallisation and cooling of large volumes of relatively hot magma (approximately 870–940 °C, based on Fe–Ti-oxide and plagioclase-melt geothermometry) at shallow crustal levels (< 1 kbar), involving release of latent heat of crystallisation, would have caused steep thermal gradients, thus favouring the instauration of vigorous hydrothermal circulation, and creating the ideal conditions for ore formation. As an indication of the link between VHMS deposition and volcanic heat, proximity to volcanic centres, as indicated by the presence of lava domes, and location in high-heat flux zones seems to be an important common feature of VHMS across geological time (de Ronde et al. 2014).

In particular, the Kuroko of the Hokuroku district is commonly, although not ubiquitously, associated with felsic domes, and is hosted in volcanic breccia on the slopes of these domes or in deposits overlying the domes (Horikoshi 1969; Sato 1977). As demonstrated in our study, these domes likely represent batches of partly degassed magma that ascended along structures such as high angle faults bounding depressed seafloor basins, intruding previous tuff deposits. Although a genetic relationship is suggested by the common spatial association, it has been previously pointed out that these domes, which average 300 m in diameter, are not large enough to provide enough metals or heat for the formation of the deposits (Sato 1977). Therefore, the domes occurred together with the mineralisation, but were not the cause of initial ore deposition. Instead, faults would have allowed both extrusion of the domes and transfer of hot metal-bearing fluids towards the surface, forming a high-heat environment.

Conclusion

The study of textures and chemical analyses of rocks, minerals and melt inclusions provides new important information on the evolution, emplacement and crystallisation conditions of volcanic rocks hosting the Kuroko deposits of north–east Japan. The main points can be summarised as follows.

  • Homogeneous plagioclase textures and compositions indicate large, compositionally homogeneous magma reservoirs located at very shallow conditions (< 1 kbar, approximately < 4 km of crustal depth) that evolved to very high SiO2 compositions.

  • Major and trace element compositions of high-SiO2 rhyolitic melt inclusions suggest magma crystallisation at near-thermal minimum (cotectic) in the granitic system, which occurred mostly through crystal fractionation accompanied by periodic volatile saturation, fluid exsolution and vesiculation.

  • Low H2O estimates based on plagioclase-melt equilibration (approximately 2 ± 0.5 wt%) are also compatible with shallow crystallisation and magma storage conditions (around 0.5 kbar), in agreement with the melt normative compositions.

  • Fe–Ti-oxide compositions indicate temperatures around 870–890 °C and relatively oxidised conditions (FMQ + 1.5), and plagioclase-melt compositions indicate temperatures of 880–940 °C.

  • Melt inclusion analyses indicate that some metals were retained in the melt during magma evolution and degassing (Zn, in particular), whereas others (e.g. Cu and Sn) were scavenged by exsolving fluids. This contrasting behaviour of metals helps explain the fractionation of Zn and Cu between black and yellow ores in Kuroko deposits.

  • Magma storage at (very) shallow conditions, and release of degassed magma and fluids along faults, would have promoted heat transfer towards the surface and fluid focussing, and induced vigorous hydrothermal circulation, a necessary condition for ore deposition.