Introduction

Quenching is expected to occur when a silicate melt comes in contact with a much cooler substrate. There is, however, another form of quenching that is near-isothermal and that is triggered by degassing and fluid exsolution (Donaldson 1974; Applegarth et al. 2013; Audétat and Edmonds 2020). We call this form of quenching “degassing-quenching”. When melts are saturated with volatile compounds and lose these volatiles, their liquidus temperatures are raised while the system temperature remains near-constant (Donaldson 1974). Following volatile exsolution, the melt is supersaturated with crystalline phases and supercooled. The volatile compound most effective in triggering degassing-quenching is H2O because H2O has the greatest influence on the liquidus of silicate melts, notably at elevated pressure.

In Cyprus we have a convincing case of degassing-quenching. In the Upper Pillow Lavas (UPL) of Kalavassos in the northeastern Troodos ophiolite, fresh lavas exist that are exceptionally H2O rich. The distinctive and unusual feature of Kalavassos is that prior to or during eruption, the lavas exsolved a fluid phase. Precipitates of the exsolved fluid are preserved alongside the basaltic glasses (Ballhaus et al. 2015; Fonseca et al. 2017). This is an unusual case; normally, exsolved volatiles are ephemeral components that leave the melt before or during emplacement (cf. Plank et al. 2013). Thus, for the first time, we may be able to establish a clear link between H2O saturated basaltic melts and their fluid phases, as well as characterise the nature of hydrous magmatic fluids.

In this paper we quantify the concept of degassing-quenching. We report major and trace element compositions and H2O contents of basaltic andesite and boninitic glass orbicules that coexist with their fluid precipitates. We document a diverse collection of quench crystals and fluid precipitates that formed following degassing-quenching. One quench clinopyroxene was sectioned with Focussed Ion Beam (FIB) techniques and examined by Transmission Electron Microscopy (TEM) to explore the consequences of rapid crystal growth on the nano-crystallography. Toward the end of the paper we speculate on the significance of degassing-quenching for spinifex textures in komatiites.

Samples

Figure 1 illustrates pockets of orbicular basalts and fluid precipitates, plus a scheme as to how degassing-quenching may work. The melts form near-perfectly spherical to ovoid-shaped glass orbicules (Fig. 1a) that are concentrated in the cores of fresh pillows and lava tubes of the UPL of Kalavassos (N34°48.101 E33°15.950′). Liquidus phases were olivine and subordinate clinopyroxene, together around 10 vol.%, plus fresh glass. The spherical structures are thought to have formed when the melts segregated a hydrous fluid phase and sought minimal surface-to-volume ratios against the fluid matrix (Ballhaus et al. 2015). The textures are comparable to emulsions of oil in water and vice versa.

Fig. 1
figure 1

Emulsions of hydrous basaltic melt and a fluid phase, to illustrate the concept of degassing-quenching. a Basaltic glass spherules within a pillow core. b Former fluid phase, now crystalline fluid solute (fs) spherules. The spherules consist of micro-spinifex clinopyroxene needles, minor K-feldspar, chlorite, clay minerals (± palagonite), and Fe-hydroxides or hematite (red specs). Open spaces between fluid solute spherules were originally (at high T) filled with later-stage fluids that were set free from the solute as the latter crystallised, then lost by surface weathering (→ porosity). Both images from the UPL of Kalavassos, Cyprus (N34°48.181′ E33°15.978′). c Concept of degassing-quenching in P–T-H2O space. Basaltic melt with 3.9 ± 0.4 wt.% H2O (Plank et al. 2013) at the H2O saturated liquidus (ca. 1080 °C) exsolves an H2O rich fluid phase and becomes supersaturated with crystalline phases. Liquidus depression as a function of H2O content quantified with Eq. (17) of Médard and Grove (2008), H2O solubility at fluid saturation from Dixon et al. (1995), dry liquidus temperature (1180 °C) approximated with the MgO-in-melt thermometer (average 8 wt.% MgO) of Helz and Thornber (1987). For further details see Matveev and Ballhaus (2002) and Ballhaus et al. (2015)

Full size image

Locally, pockets of exsolved and solidified fluid solute are preserved. At the UPL locality of Kalavassos decimeter-sized enclaves exist adjacent to the glass orbicules in which light-grey spherules are concentrated (Fig. 1b). Some solute spherules are radial in structure. Phases are elongate clinopyroxene in micro-spinifex texture and minor K-feldspar. Between the needles, selvages of chlorite and clay minerals (likely palagonite) are prevalent but are too fine (nm scale) to be fully quantified (cf. Fonseca et al. 2017 for more information). Glass is not preserved if it ever existed. We interpret the spherules as solute precipitated from a magmatic fluid phase that segregated from the hydrous melts (Fig. 1a), then collected in discrete pockets. A notable feature is the high porosity between the solute spherules. We assume that originally, the pore space was filled at sub-magmatic temperature with an H2O enriched derivative fluid, released when the solute crystallised, then cleared out by surface weathering.

Figure 1c illustrates the concept of degassing-quenching (Donaldson 1974). It displays the dry (solid line) and the H2O saturated liquidi (dotted line), of a tholeiitic basalt with an MgO content of 8 wt.% MgO. The pressure effect on the dry liquidus is schematic. The liquidus depression relative to the dry liquidus of the basalt was calculated with the formalism of Médard and Grove (2008), and the H2O solubility at fluid saturation is from Dixon et al. (1995). The red arrow labelled ∆T between the dry and H2O saturated liquidi illustrates the degree of undercooling (supersaturation) of a melt at ~ 150 MPa that originally contained 3.9 wt.% H2O (Plank et al. 2013). This ∆T value is a maximum because melts emplaced on the seafloor cannot lose all their H2O during fluid exsolution. In the ensuing discussion we will adapt Fig. 1c for the Kalavassos situation.

Methods

Phase analysis

Crystalline phases, glasses, and fluid solutes were analyzed for major oxides with a JEOL JXA 8200 electron probe at the University of Bonn at 15 kV and 20 nA using natural and synthetic materials as standards and the Juan de Fuca Ridge glass VG-2 (Jarosewich et al. 1980) as secondary reference material. Crystals were analysed in spot mode, glasses with beams defocussed to 5 to 10 µm, and the much coarser grained solute precipitates with 30 µm beams. Two glass orbicules (zy-9 und zy-15) were analysed at the University of Göttingen with the Field Emission Electron Probe JEOL JXA-iHP200F for dissolved S2−. Sulfur was calibrated on ZnS. The glasses were measured using three spectrometers simultaneously with PETH and PETL crystals at 15 kV and 20 nA for 60 s, each on peak and background. Crystal phases in the glasses and fluid spherules were imaged with a Zeiss Gemini Sigma 300 VP Field Emission Scanning Electron Microscope at the University of Cologne. One clinopyroxene quench crystal was sectioned twice with a Thermo Scientific Dual Beam Helios G4 UC FIB-SEM ± perpendicular to (010) and (001), then investigated for its nanostructures with a FEI Tecnai G2 F20 X-Twin and a Thermo Scientific THEMIS Z Transmission Electron Microscope (TEM) at the GFZ in Potsdam (cf. Wirth 2009).

Quantification of H2O with FTIR

To quantify the H2O contents of the glasses, four orbicules were analysed with FTIR spectroscopy (Bruker IFS 120 HR) at the Geoinstitut in Bayreuth. The spectroscope is equipped with a Bruker infrared microscope IRscope I, a tungsten light source, a CaF2 beamsplitter, and a Mercury-Cadmium-Telluride (MCT) narrow range detector. Measurements were carried out on doubly polished thin sections with thicknesses ranging from 320 to 370 µm. The OH contents were determined from the 4500 cm−1 band and the H2O contents from the 5200 cm−1 band (extinction coefficients from Shishkina et al. 2010). The results were added up to total H2O contents. The spectral resolution was 4 cm−1. For each spectrum, 200 scans were collected from spots with sizes 200 μm in diameter. The method is described in detail by Shishkina et al. (2010).

Trace element analysis by quadrupole ICP-MS

Eight glass orbicules (Fig. 1a) and 7 solute spherules (Fig. 1b) from the Kalavassos locality were selected for trace element analysis. In addition, two larger glass orbs 5 cm in diameter, from Kapileio (N34°49.755′ E32°57.850′), were cut into approximately 10 mm thick slices to take 4 and 7 glassy samples from the interior, respectively of each orbicule. The Kapileio locality is genetically similar to Kalavassos (cf. Ballhaus et al. 2015). Before preparation, samples were ultrasonicated in ethanol for 20 min, then rinsed three times with de-ionized water, dried at 100 °C, and crushed to < 1 mm fragments in an agate mortar. For the trace element analysis, 30–150 mg of these fragments and 50 mg of the basaltic reference material BVA-1 (GeoPT proficiency test 2022, www.geoanalyst.org/geopt/) were weighed into Savillex® PFA vials. Samples were then doped with an internal Re-Rh standard solution and its weight recorded, in order to correct for measurement drift as well as imprecise internal standard addition and dilution (Braukmüller et al. 2018).

The sample powders were digested in 3 mL concentrated HNO3 and HF (1:1) on a hotplate at 120 °C for 48 h. After the solutions were dried down to incipient dryness, 5 ml of concentrated HNO3 were added and the solutions dried down again. The nitric acid dry-down step serves to remove insoluble fluorides and was repeated twice. The samples were then ultrasonicated. Because fluoride precipitates were still observed, the digestion/dry down steps were repeated twice to ensure that any residual material was dissolved. Following complete dry-down, the samples were dissolved in 1.5 ml 2.8 N HNO3 and visually inspected for suspended solids (fluorides) by drawing up the solution with a transparent pipette.

Trace element analysis was carried out on sample solutions diluted 4000 and 8000 fold in 0.28 N HNO3. Trace elements were analysed with a Thermo ScientificTM iCapTM Q quadrupole ICP-MS at the Institut für Geologie und Mineralogie at the Universität zu Köln, using a Peltier-cooled cyclonic PFA spray chamber and a self-aspirating MicroFlow PFA nebulizer with an uptake rate of about 100 µl/min. All sample solutions were measured multiple times within two measurement sessions. Instrument parameters were 100,000 and 130,000 counts per second (cps) for 115In (1 ng/ml In), Ce+/CeO+ of 1.2 and 1.7%, and Ba2+/Ba+ of 3.6%, respectively, in the standard measurement (STD) mode. In the kinetic energy discrimination (KED) mode, the intensities achieved were 19,000 and 24,000 cps for 1 ng/ml 59Co and Ce+/CeO+ rates of 0.7 and 0.6%, respectively.

External calibration utilized solutions prepared from the reference materials BIR-1, BHVO-2, BCR-2, JB-2, and AGV-2 (Jochum et al. 2005, 2016). In general, our intermediate precision is better than 6% RSD.

Chemical compositions

Representative major element compositions of glass orbicules (Fig. 1a) and solute spherules (Fig. 1b) are summarised in Table 1. The analyses are averages of ~ 20 electron probe microanalyses per sample performed with defocussed electron beams. Silica and MgO contents (averages 54.3 ± 1.2 and 7.6 ± 1 wt.%, respectively) qualify the glasses as basaltic andesites and high-Ca boninites (MgO > 8 wt.%; Crawford et al. 1989). Molar Mg# (0.63–0.7) and coexisting olivines (Fo85 to Fo89) are primitive to sub-primitive. Ferric iron contents of the glasses (Table 1) were calculated assuming the KDFe-Mgol−melt was 0.3. Relative to the fayalite-magnetite-quartz (FMQ) buffer, oxygen fugacities (fO2) are moderately oxidised and fall around FMQ – 0.3 ± 0.7 (1σ) (Kress and Carmichael 1991).

Table 1 Major oxides of glass orbicules and fluid solute
Full size table

The H2O contents measured on four glass orbicules (Methods) are quantified at 2.5 ± 0.25 wt.% (1σ, Table 1). They fall at the upper end of the H2O range (1.7–2.6 wt.%) reported for Troodos UPL (Muenow et al. 1990; Portnyagin et al. 1997; Sobolev and Chaussidon 1996; Woelki et al. 2020). Fonseca et al. (2017) analysed H2O of Kalavassos samples with micro-Raman and found concentrations of 2.4 ± 0.1 wt.%, near-identical to our FTIR analyses. Sulfur contents of two representative glass orbicules quantified with the JEOL JXA-iHP200F on SKα1 are low at 250 ± 13 ppm (1σ). Although not all orbicules were analyzed for H2O and S, we think the concentrations are representative for all glass orbicules reported in Table 1.

Solute compositions (Table 1) are more variable than those of the glasses. Much of the variability is real; variations among individual probe analyses are significantly less than variabilities among individual spheres. Silica ranges around 53.2 ± 3.3 wt.% (1σ) and is not significantly higher but more variable than the SiO2 of the glasses. Alumina contents (21.5 ± 1.2 wt.%) are high, as are K2O (2.5 ± 1.7 wt.%) and K2O/Na2O (2.1 compared to 0.09 of the glass orbicules). We suspect that significant Na2O and possibly SiO2 were mobilized into the open pore space between the spherules (Fig. 1b) and are now lost. As suggested above, the pore space may have been filled originally with late-stage fluids that were released when the solute crystallised.

In Table 2 we summarize trace element contents, and Fig. 2 displays the mantle normalised (Palme and O’Neill 2014) element patterns of glass orbicules (Fig. 2a) and fluid solute spherules (Fig. 2b). Spectra that are nearly identical are combined for clarity. Because the samples in Table 1 were all embedded in epoxy resin, hence no longer available for destructive trace element analysis, the trace element contents were determined on spherules that were mineralogically identical to those in Table 1. Two orbicules are from Kapileio, a locality genetically similar to Kalavassos (Methods). The elements are ordered in Fig. 2 from left to right with increasing compatibility at partial melting (Hofmann 1997).

Table 2 Trace element concentrations of glass orbicules and solute spherules
Full size table
Fig. 2
figure 2

Primitive mantle-normalised (Palme and O'Neill 2014) trace element patterns of nine glass orbicules (a) and seven fluid solute spherules (b). Order of elements with increasing compatibility to the right (Hofmann 1997). The lighter-coloured patterns are highlighted because they are more fractionated (see text). Distinct anomalies marked on the graphs

Full size image

Qualitatively, there are many similarities between glasses and fluid precipitates. Both melt and solute compositions show marked enrichments in the fluid mobile elements K, Rb, Pb, and Sr. Without the fluid mobile elements, a slight U-shape is recognized in both the melt and the fluid patterns, which Cameron (1985), König et al. (2008), Golowin et al. (2017), and Fonseca et al. (2017) also noted. That pattern seems typical for both the UPL lavas and boninitic cumulates from Cyprus (Shen et al. 2020) as well as for boninites elsewhere (Parman et al. 2001).

Nine glass orbicules are near-identical in trace element contents and are summarised as one pattern shown in dark green. Among those are two larger orbicules (Zy32 and ZY-E) from Kapileio, i.e., the averages of 4 and 7 analyses, respectively (cf. Methods). All those spectra show moderately fractionated La/LuN of 0.55 ± 0.05. One glass orbicule (zy-22) is highlighted in pale-green because it has a more pronounced fluid signature, a larger positive Eu anomaly, a lower La/LuN of 0.44, and appears to be more fractionated. The Eu anomaly may indicate a larger enrichment in plagioclase component whose crystallisation was suppressed by high pH2O (cf. Crawford et al. 1987). The Zr/Hf and Y/Ho ratios (29 ± 0.7 and 27 ± 1 respectively) fall in the CHARAC field defined by Bau (1996) and are typical for high temperature magmatic rocks.

The fluid solutes also comprise two populations that differ with respect to fractionation degree (Fig. 2b). Common to all solute compositions are marked positive anomalies in the mobile elements Rb, K, Pb, and Sr resembling the fluid signatures noted in the glass orbicules. Four samples (in red) are in many respects similar to the glass orbicules and have surprisingly high La/LuN of 0.74. Three fluid solute samples (in pale red) are more fractionated and show depleted La/LuN of 0.15 ± 0.01. Both solute fractions have near-identical Y/Ho (25 ± 0.8) but differ from the glasses with respect to Zr/Hf (22 ± 0.6 vs. 29 ± 0.4).

Quench crystals in basaltic glasses and fluid precipitates

In Fig. 3 we compile a selection of crystal phases and crystal shapes that crystallised following degassing quenching. Olivines range from euhedral to distinctly hopper-shaped (Fig. 3a and b). The cavities that are produced by rapid growth (cf. Donaldson 1976; Faure et al. 2003; Faure and Schiano 2005) are filled with fluid solute, in Fig. 3a and b by chlorite and clay minerals, not unlike quench olivines illustrated by Belgrano et al. (2021). Many glass regions around olivine quench skeletons show concentric contraction cracks (Fig. 3c) decorated with nanoscale fluid or vapor inclusions. The textural impression is that cores of olivine were already stable when fluid saturation set in, but then olivine continued to grow rapidly with increasing ∆T (Fig. 1c), forming hoppers and trapping fluid precipitates in their cavities.

Fig. 3
figure 3

Phases that precipitated following degassing-quenching (cf. Fig. 1c). a Hopper-shaped olivine with hydrous fluid solute trapped in its cavities (fs, largely chlorite and palagonite). Intimate association of olivine with fluid solute indicates that this olivine grew in the presence of an exsolving fluid phase. BSE image. b Olivine skeleton within fluid solute (fs); that the solute appears to outline a euhedral olivine shape does not mean olivine was replaced by chlorite, more likely this a 2D effect. BSE image. c Olivine edge rimmed by palagonite; note the concentric shrinking cracks decorated by dark inclusions, possibly voids or vapor. BSE image. d Clinopyroxene quench crystal in hydrous glass, section ± normal to (001); positions of FIB sections in red. BSE image. e Euhedral equilibrium clinopyroxene with darkish relict of pigeonite in its center. BSE image. f Bundle of clinopyroxene quench crystals in glass with near-euhedral outlines. BSE image. g Fluid solute with roundish cavities rimmed by chlorite; the fine grained matrix (former fluid solute, fs) consists of micro-spinifex clinopyroxene with nanometer wide darkish chlorite lamellae, some K feldspar, and some clay minerals (inset, scale bar 4 µm). The oval structures with central cavities formed when the solute crystallised, concentrating latest-stage exsolved H2O into discrete cavities (cf. porosity in Fig. 1c). Secondary electron image

Full size image

Many clinopyroxenes form dendrites and rather extreme feathery quench forms (Fig. 3d and e). We find no fluid solute attached to clinopyroxene dendrites nor to clinopyroxene equilibrium crystals (Fig. 3f). All crystals are completely immersed in clean basaltic glass. Hence we assume that clinopyroxene was stabilized after olivine, at a time when fluid was exsolved and had escaped the system (cf. Figure 1b), when ∆T had reached its maximum (Fig. 1c). The first pyroxene phase may have been pigeonite, now largely resorbed, as some euhedral clinopyroxene crystals carry relicts of pigeonite (Fig. 3f).

A striking observation is that many clinopyroxene dendrites (cf. Fig. 3d and f) could form euhedral shapes. The fictitious “crystal faces”, that is, interpolated boundaries of the dendrites to the glass, are occupied only by seemingly isolated crystal satellites. They nevertheless trace the nominal crystal faces of the monoclinic shape almost perfectly. The fluid solute spherules are also largely composed of micro-spinifex clinopyroxene (Fig. 3g) not unlike spinifex pyroxene in komatiites (Faure et al. 2006) albeit on a much smaller scale. Some spherules are dominated by chlorite and palagonite. The cavities in the fluid precipitate (Fig. 3g) have rims of chlorite around central voids. We think they mark locations where latest-stage H2O rich fluids was collected (cf. pale red spectra in Fig. 2b), released when the spinifex clinopyroxenes grew, then lost during cooling. On a microscale they could correspond to the macroscale porosity of Fig. 1b.

Clinopyroxene dendrites grew so fast that equilibrium partitioning broke down (Grove and Bence 1979). In Fig. 4 we illustrate element maps of a dendrite acquired with an electron beam 1 µm in diameter with 2 µm steps. Virtually all element fractionations during crystal growth are internal, that is, remained confined to the melt fractions trapped inside the dendrite. Compositional boundary layers to the outside glass are narrow, no wider than 5 µm but compositional gradients to the main glass are steep. The bulk composition of the skeleton determined with an electron beam defocussed to cover much of the skeleton's surface is not systematically different from the composition of the matrix glass. The glass interstitial to the central “spine”, best visible in the BSE and Si images of Fig. 4, is quite enriched in silica and depleted in Fe, Mg, and Ca, but too small for contamination free probe analysis.

Fig. 4
figure 4

Element maps of a skeletal clinopyroxene (BSE image to the left; Fonseca et al. 2017). The bulk composition of the skeleton determined by defocussed electron beam is near-identical to the composition of the matrix glass, suggesting that element fractionation during crystal growth was largely confined to within the skeleton. The glass between the branches (dark in the BSE image and best visible in the Ca image) is SiO2-Al2O3-enriched but depleted in Mg, Fe, and Ca. Boundary layers around the skeleton are no wider than 5 µm. Color coding: red, magenta, and white represent highest concentrations, black and dark blue lowest

Full size image

Figure 5 summarises clinopyroxene compositions in the pyroxene quadrilateral (quad). The phase diagram displays both equilibrium shapes (in blue) and metastable dendrites (in red). The quench pyroxenes are highly sub-calcic and fall in the two-pyroxene miscibility gap. Their positions in the quad are due to high (disequilibrium) concentrations in Al2O3, up to 12.8 wt.%, calculated as calcium tschermaks (CaTs) component and subtracted from wollastonite (cf. Lindsley 1983). We are aware that such compositions do not belong in the quad; note though that the quench pyroxenes carry exsolved nanoscale pigeonite exsolutions, to be detailed in Fig. 7. If their compositions could be resolved, pairs of augite and pigeonite would result corresponding to the schematic grey tie lines in Fig. 5.

Fig. 5
figure 5

Section of the pyroxene quadrilateral at 0.1 MPa (cf. Lindsley 1983) to show clinopyroxene compositions from the melt globules in Fig. 1a. In blue, euhedral (equilibrium) clinopyroxenes with ± 3 wt.% Al2O3 (cf. Fig. 2e), in red, skeletal (metastable) clinopyroxenes with > 6 and up 12.8 wt.% Al2O3 (calculated as CaTs molecule). Metastable clinopyroxenes have exsolved nanoscale pigeonite lamellae. The light grey dashed tie lines symbolise coexisting sub-calcic augite and pigeonite at 1100 °C that would have resulted, had it been possible to resolve analytically endmember compositions of the exsolution lamellae (cf. Fig. 7)

Full size image

Structural analysis with TEM

Two FIB sections were placed over the dendritic clinopyroxene crystal in Fig. 3d then examined by TEM. The goal was to quantify the orientation of the “satellite” crystallites of the dendrite. The question was if the crystallites are crystallographically homogeneous and connected in 3D. How did the crystals imaged in Figs. 3d and 4 manage to form pseudo-euhedral shapes when the pseudo-crystal faces are occupied only by apparently isolated clinopyroxene satellites? If the satellites are connected in 3D, how large is the crystallographic long-range order within the quench crystal?

We ablated stepwise the dendrite in Fig. 3d with a Ga+ beam using the Dual Beam Helios G4 UC FIB-SEM. Every 40 nm a BSE image was acquired (Fig. 6). We show eight slices each ~ 400 nm apart. The sequence suggests that all clinopyroxene satellites—apparently isolated in Fig. 2d–form a coherent skeleton in 3D. Indeed, the monoclinic crystal faces inferred from Figs. 3d and 4 must be highly porous, not unlike a Swiss cheese.

Fig. 6
figure 6

Sequential BSE images of the clinopyroxene dendrite in Fig. 2d. The sequence illustrates that crystallites are connected in 3D. Images produced by stepwise ablation with a Ga+ beam. Distances among the eight individual images (a–h) ~ 400 nm each. The central "spine" (x) is oriented approximately in (001) direction

Full size image

Figure 7 shows three bright-field (BF) images of several clinopyroxene crystallites that in 2D appear to be isolated. The crystallites are too small for EPMA quantification but we assume they are similar in composition to the subcalcic aluminous pyroxenes in Fig. 5 (red symbols). All crystallites carry narrow exsolutions of pigeonite and augite oriented along (001) and marked as red lines. The widths of the exsolutions lamellae are ~ 2 nm with some irregularities in the spacing (cf. Busek et al. 1980), used later to infer on the cooling rates under which the dendrite grew. We suggest they are spinodal exsolutions (cf. Grove 1982). Not only in Fig. 7; no matter where in the FIB section we measure the orientations of the exsolutions, all have an orientation relative to the vertical axis of the images of 27 ± 2°. The small variation is another indication that all segments of the skeleton in Fig. 3d are connected in 3D and that crystallographic continuity is preserved.

Fig. 7
figure 7

Brightfield images of seemingly isolated clinopyroxene segments of the dendrite in Fig. 3d (FIB # 7049). The red lines illustrate orientations of pigeonite exsolution lamellae from subcalcic aluminous augite, in this image and in all others of FIB# 7049 identical in orientation within ca. 2° (visual estimate)

Full size image

Figure 8 shows clinopyroxene offshoots ~ 50 nm in size grown onto the surfaces of larger crystallites; the inset (Fig. 8a) illustrates an overview in darkfield (DF) mode. The offshoots are seemingly isolated but that is a 2D effect since the FIB section is only around 80 nm thick. The offshoots carry exsolutions within ± 2° parallel to the exsolutions in the coarser crystallites (Fig. 8b). From one offshoot we present an HREM image and an FFT (Fig. 8c and d). The lattice planes (010) and ( – 111) lie in the exsolution plane, as do the zone axes [001] and [ – 101]. The orientation is thus clearly defined. Judging from the fine scale we suggest the lamellae are spinodal exsolutions of augite and pigeonite from sub-calcic clinopyroxene bulk compositions (cf. Fig. 5, in red). Which structures correspond to augite or pigeonite in the FFT cannot be determined since the C2/c (aug) and C21/c (pig) space groups and the unit cell parameters are too similar. It can be ruled though out that the lamellae are twins or that the exsolutions contain orthopyroxene (P21/c). Thus, they must have exsolved at a temperature when pigeonite was stable, which limits the temperature to ± 1100 °C (cf. grey tie lines in Fig. 5).

Fig. 8
figure 8

Clinopyroxene protrusions grown onto surfaces of larger clinopxroxene crystallites. a HAADF overview. b Crystallites in BF mode seemingly unconnected to the clinopxroxene dendrite but in reality (3D) crystallographically homogeneous. c and d the respective FFTs with unit cells marked in yellow and orange. The (010) and ( – 111) planes and the zone axes [001] and [ – 101] are parallel

Full size image

Figure 9 shows the opposite side of Fig. 8, and for orientation purposes it repeats the inset of Fig. 8a. Some offshoots carry ultrafine spinodal exsolutions with exactly the same orientations and widths as in Figs. 7 and 8. The HREM 03 and the corresponding FFT illustrate that both analyzed offshoots have identical structure. All offshoots show nanoscale contact with glass. Note that the glass marked in Fig. 9c is not the basaltic matrix glass in which the pyroxene skeletons are embedded (cf. Fig. 3d); it is the residual glass that formed by internal fractionation within the skeleton, i.e., the dark material between the dendritic branches in Fig. 3d. The composition of this glass is not precisely known but SiO2 and Al2O3 are highly enriched (cf. Fig. 4).

Fig. 9
figure 9

Opposite side of the clinopyroxene crystallite in Fig. 8. a HAADF image as overview. b BF image with exsolution lamellae marked as red streaks. In the HREM image (c) and the FFT (d) one crystal structure detected, possibly that of augite since augite is the major component in the quench clinopyroxenes; note the contact with glass inside the quench clinopyroxene (cf. Figure 3d). Unit cell marked on the FFT in (d)

Full size image

To gain further insight into the orientation of the skeleton segments, micro-diffractions (XRMD) were performed on 15 seemingly isolated crystallites (Figs. 10 and 11). The measurement positions are marked with yellow circles in the DF and BF images to the left. The question was whether and to what extent crystallographic homogeneity is preserved, given that the orientations of the exsolution lamellae inserted with the naked eye (cf. Fig. 7) are not a fully quantitative criterion. The red circles in the diffractograms of Figs. 10 and 11 (to the right) mark the positions of the transmitted (unscattered) beam indexed as 000, and the yellow dots the tilt of the zone axis [001] relative to the unscattered beam. The maximum tilt angle is 2.8°, similar to variations among the spinodal exsolution orientations (± 2°). We think the figures give a good impression of how far the long-range order goes within the clinopyroxene skeleton investigated.

Fig. 10
figure 10

Left: Dark-field (DF) image of crystallites of the dendrite in Fig. 3d. Positions of microdiffractions indicated by yellow circles. To the right, microdiffractions of the crystal satellites marked in the DF image to the left. The yellow dots indicate the deviation of the zone axis of the diffraction pattern from the optical axis (red dots) of the TEM

Full size image
Fig. 11
figure 11

Bright-field image (BF08) of the clinopyroxene satellites shown in Fig. 8, diffraction positions numbered by yellow circles. To the right, micro-diffraction patterns; the yellow dots indicate the deviation of the zone axis of the diffraction pattern from the optical axis (red dots) of the TEM. The main grain (1) is oriented along the [001] direction of a pigeonite crystal whereas the small grains are oriented close to [101] and are slightly rotated along [010]. The tilt angles are listed on the figures

Full size image

Sulfide nano-inclusions

An intriguing side aspect is that glass pools trapped inside the pyroxene dendrite carry FeS nanoparticles (Fig. 12a). The inclusions appear to be confined to the SiO2 enriched internal glasses wedged between the dendrite branches (cf. Fig. 4). No nanosulfides were found in the ± primitive basaltic glasses outside the dendrite (cf. the HAADF image in Fig. 12b). The FeS nanophases are crystalline (Fig. 12c and d). We note that all sulfide nanocrystals imaged are attached to fluid phases. Figure 12e illustrates three sequential images of an FeS inclusion that seems to be embedded in a fluid (possibly vapor) phase. The presence of a fluid is inferred because the bubbles (in dark) change within seconds under the electron beam their shapes and their positions. Electron beam heating causes local density fluctuations resulting in variable electron absorption during imaging. Hence we identify in the Kalavassos samples fluids at four orders in magnitude in size: the mm-sized porosity around the solute spheres (Fig. 1b), the voids rimmed by chlorite in the solute precipitates (Fig. 3g), the micron-sized fluid (or vapor) inclusions decorating cracks in the glass orbicules (Fig. 3c), and nanosized fluid inclusions attached to the FeS nanocrystals (Fig. 12a).

Fig. 12
figure 12

Sulfide inclusions in glass inside the pyroxene skeleton (cf. Figure 3d). a HAADF image, bright spots are FeS nano-inclusions, possibly FeS aggregates, with fluid phase attached. The FeS nanophases are located in the glass interstitial to the dendrite branches (cf. dark material in Fig. 3d). b HAADF image of the basaltic glass surrounding the skeleton, no FeS inclusions recorded. c HREM image of an FeS nano-inclusion in glass outlined by yellow dashes, and in (d) the indexed FFT. e Three consecutive BF images of an FeS nano-inclusion (bright) in glass; attached to FeS a fluid or vapor bubble (in black) that changes its position with time under the electron beam

Full size image

Since TEM did not identify FeS inclusions in the basaltic matrix glass, the question arose whether the glasses in Table 1 were S undersaturated. We analyzed two glass orbicules (zy-9 and zy-15 in Table 1) for dissolved S2− using the FE Electron Probe at the University of Göttingen (Methods). Sulfur concentrations measured on the Kα1 line of S2− are low and amount to 250 ± 13 ppm. No measurements were made on the sulfate line because the relative fO2 of the glass orbicules (around FMQ) rules out S as SO42− (Jugo 2009; O’Neill 2022; O’Neill and Mavrogenes 2022).

If 250 ± 13 ppm S are representative for all the glass orbicules in Fig. 1a, the glasses are by factor 3.5 to 4 undersaturated with respect to FeS saturation. We calculate for the glasses in Table 1 nominal S contents at sulfur saturation (SCSS) of 840 to 1040 ppm (O’Neill 2022). Thus, the fact that the FeS nanocrystals are confined to the silica rich glasses inside the clinopyroxene skeletons can be explained either by accumulation of S2− during rapid crystal growth, since S2− is incompatible with clinopyroxene (Callegaro et al. 2020), or they reflect the high SiO2 and low NBO/T of the glasses. Increasing SiO2 lowers SCSS (Haughton et al. 1974; Shima and Naldrett 1975; O’Neill 2022) and triggers sulfide saturation.

Discussion

Chemical compositions

The melt compositions in Table 1 are classified as basaltic andesites (< 8 wt.% MgO) and high-Ca boninites (> 8 wt.% MgO, average CaO 11.7 ± 0.28 wt.%) (Crawford et al. 1989). The sequence of events was crystallisation of near-primitive olivine and chromite → fluid saturation → (pigeonite →) high-Ca clinopyroxene nucleation. Pigeonite is set in brackets because it is a rare relict only in euhedral clinopyroxenes (cf. Fig. 3f). Compositions with molar Mg/(Mg + Fe) > 0.65 and olivine compositions of Fo85-89 are identified as near-primitive. Some glasses with SiO2 > 54 wt.% could be slightly fractionated either by conventional olivine/pigeonite fractionation and/or by the exsolution of a fluid phase. The mantle source to the UPL may have been a moderately depleted, yet clinopyroxene saturated peridotitic mantle wedge, hydrated and brought to second-stage partial melting by the infiltration of dehydration fluids from an underlying slab (Crawford et al. 1989; van der Laan et al. 1989; Sobolev and Chaussidon 1996; Golowin et al. 2017).

There can be little doubt that the glass orbicules and the solute spherules have a common parentage. The trace element patterns in Fig. 2 are qualitatively identical. Fluid saturation is “dated” to a time when olivine crystallized (cf. Fig. 3a–c) because skeletal olivines are often lined by chlorite and palagonite, but it was completed when clinopyroxene nucleated. We identify two solute compositions that differ in fractionation degree (Fig. 2b) but think they form part of a continuum of compositions exsolved over a range of temperature but not fully represented by samples (cf. Audétat and Edmonds 2020). We note that the fluids lack a fluid signature as defined by Bau (1996). The Y/Ho and Zr/Hf ratios are hardly distinguishable from those of the melt orbicules. We attribute this to the high temperatures at which the fluids segregated. The high variability among the REE remains speculative since data on REE solubility in magmatic fluids are contradictory. Audétat (2019) reported for Ce a Dfluid/melt of ~ 0.02 while Banks et al. (1994) documented elevated REE in fluids rich in halogens. Whatever the solution; if late-stage fluids were lost as indicated by the porosity in Fig. 1b, the depletion in LREE is easy to reconcile whether fluids were enriched or depleted in REE. The Lu contents, near-identical to those of the glass orbicules, should reflect modal clinopyroxene abundance since the HREE are more compatible with clinopyroxene than the LREE (Hauri et al. 1994).

Seawater probably did not contribute significantly to the H2O of the fluid solutes or the glasses. The δ11B (down to—10‰) of Kalavassos orbicules are too negative for seawater. Fonseca et al. (2017) argued based on B isotope ratios that the fluids were mantle derived, that they came from a source contaminated prior to partial melting by a sedimentary component. Physical constraints also argue against a seawater contribution: seawater would have had difficulty infiltrating the orbicular units against the pH2O imposed by the H2O dissolved in the melts.

Unfortunately, bulk fluid compositions cannot be deduced since the late-stage fluid fraction that must have filled the pore spaces in Fig. 1b is lost. Nonetheless, it can be stated that hydrous magmatic fluids at elevated pressure are rich in oxide solute. If we could sample them directly at pressure and temperature, they would rather correspond to viscous gels than to fluid phases (Ballhaus et al. 2015).

Degassing-quenching

The H2O contents of the glasses permit the quantification of “degassing-quenching”. The H2O in the glass orbicules must be residual since the glass orbicules coexist with fluid solute. The S contents of 250 ± 13 ppm are also residual as they are by factor 3.5–4 lower than the S content at sulfur saturation (SCSS; O’Neill 2022). Keppler (1999) showed that sulfur fractionates strongly into fluid phases when melts reach fluid saturation. The DSfluid−melt listed by Keppler range from 47 ± 4 (SO2) to 468 ± 32 (H2S), so the reason for the low SCSS might be S loss via exsolution of the fluid phase. Perhaps the preference of S to fluid phases explains the frequent association of sulfides with volatile rich phases (cf. Ballhaus and Stumpfl 1986; Daryl et al. 2020).

A residual H2O content of 2.5 ± 0.25 wt.% (Table 1) corresponds to a nominal pH2O of 65 ± 10 MPa (Dixon et al. 1995). This pressure corresponds to a seawater depth of 6500 ± 1000 m, a reasonable depth range for many backarc basins (Park et al. 1990) and in good agreement with water depths derived by Belgrano et al. (2021) for UPL lavas from various localities in the Troodos massif. The problem is that we do not know if the glass orbicules reached pH2O equilibrium with the overlying seawater column while they erupted and exsolved a fluid phase. Stolper (1982) showed that at 2.5 wt.% H2O (our glasses), “water” is dissolved mostly as OH (hydroxyl), less so as molecular H2O. The exsolution of a hydrous fluid thus requires a re-organization of the melt structure toward more polymerized units, and if the solidification of the orbicules to glass was faster than the attainment of pH2O equilibrium with seawater (cf. Zhang and Ni 2010) part of the H2O detected in the glass orbicules may be metastable. Hence we cannot rule out that the emplacement depth derived above is an overestimate.

To quantify the depression of liquidus temperature by H2O (∆T in Fig. 1c) the pre-fluid exsolution H2O contents must be known. A good guide is provided by the H2O contents of primitive island arc melts worldwide that Plank et al. (2013) reported at 3.9 ± 0.4 wt.%. Presumably, an arc setting also applies to Troodos. The UPL formed in the vicinity of a backarc spreading center or a triple junction above a wet subduction zone (cf. Pearce and Robinson 2010; Regelous et al. 2014). For a pre-eruption H2O content of 3.9 ± 0.4 wt.%, we calculate a liquidus depression ∆T relative to the dry liquidus of − 117 ± 8 °C (Fig. 1c). After subtracting the 2.5 ± 0.25 wt.% of residual H2O (Table 1), the ∆T relative to the dry liquidus (Helz and Thornber 1987) reduces to − 65 ± 10 °C, for the orbicules in Fig. 1a.

Dendritic crystals and cooling rates

The dendritic pyroxene crystals and their compositions may help derive a cooling rate ∆T/t, where t stands for time. The parameter ∆T/t is composed of the increase in liquidus temperature and entropy gain by fluid exsolution, plus conductive cooling after the melts were emplaced on the ocean floor.

Dendritic crystal morphologies result when crystals grow under conditions of strong supercooling, when growth rates are significantly faster than diffusion rates (Bryan 1972). Rapid growth enhances the formation of boundary layers between crystals and melt, depleted in all components required by the crystal to grow yet enriched in rejected solute (Fig. 4). For geometric reasons, boundary layers should be thinnest at convex kinks and corners of a crystal. Since high concentrations in rejected solute impede crystal growth, we propose that all morphological elements of a skeleton protruding into the glass should tend to grow faster than crystal faces. Hence, skeletal morphologies will result. A consequence of dendritic growth is the formation of fractal elements (Fowler et al. 1989; Meng et al. 2013) which we actually identify in the dendrite in Fig. 3d on several orders in size (Fig. 13). Rapid growth did not affect adversely the long-range order (cf. Figs. 10 and 11) but why the dendrite branches stopped growing when they reached the nominal crystal faces of monoclinic pyroxene remains unsolved.

Fig. 13
figure 13

Fractal elements in the clinopyroxene skeleton (cf. Figure 3d). a BSE image as in Fig. 3d; the main morphological element is the central spine (numbered 1) that is repeated many times on a smaller scale (red arrow). b Brightfield image of a smaller-scale segment oriented normal to the spine in Fig. 13a (numbered 2); the small offshoots (numbered 3) are all connected in 3D. c and d Darkfield images of small “twigs” and “buttons” grown ± vertically onto a branch oriented in (010), identified as fractal elements no. 4 and 5. Five orders of magnitude in size identified among the fractals, ranging from ~ 100 µm to 10 nm

Full size image

Rapid growth also influenced crystal compositions. The dendritic pyroxenes (blue symbols in Fig. 5) grew so rapidly that equilibrium partitioning broke down (cf. Grove and Bence 1979, Coish and Taylor 1979). Mollo et al. (2013) calibrated the DAl2O3cpx−melt over a wide range of ∆T/t and showed that DAl2O3cpx−melt increases significantly with increasing ∆T/t. Our most aluminous dendrite segment analysed has 12.8 wt.% Al2O3. The average Al2O3 content of the glass orbicules is 15.3 ± 0.9 wt.% (1σ; Table 1). Hence we calculate a DAl2O3cpx−melt of 0.83 ± 0.04 (Mollo et al. 2013) corresponding to a ∆T/t of about − 50 °C h−1. Note though that this value could be rather imprecise: all clinopyroxene skeletons are embedded in boundary layers enriched in rejected solute, including Al2O3 (cf. Fig. 4).

The widths of the augite-pigeonite exsolutions (Fig. 7) may provide a further estimate of the cooling rates. Our augite-pigeonite exsolutions have widths around 2 nm, i.e. an order of magnitude finer than those produced experimentally by Grove (1982) with a linear cooling rate of 150 °C h−1. A contradiction seems to arise here. A plausible solution is that natural cooling rates are not linear but rather exponential; that the initial cooling rates during incipient fluid exsolution and degassing-quenching may have been faster than Grove's experimental cooling rate of − 150 °C h−1, and much faster than cooling rates derived from metastable Al2O3.

Implications—spinifex textures in komatiites by degassing-quenching?

The appropriate targets to look for degassing-quenching are volcanic units with quench crystals that could not have been formed by conductive cooling. Degassing-quenching is a whole-sale process that does not require a local temperature gradient. Quenching rates induced by volatile loss are independent of melt volumes since ∆T/t only depends on the amount of fluid lost and the fluid loss rate, affecting large melt volumes as well as small ones.

Our study might offer a solution to the so-called spinifex paradox (Nesbitt 1971). The paradox states that spinifex textures are found inside thick komatiite flows, sometimes as deep as 10 m below the top chills where the cooling rates are no more than 0.3 to 1 °C h−1 (Turner et al. 1986; Dann 2001). Such rates are far too slow for spinifex textures to develop (Donaldson 1982; Arndt and Fowler 2004)—yet they are observed. Faure et al. (2006) and Bouquain et al. (2014) duplicated spinifex textures experimentally by performing experiments in long capsules within large thermal gradients around 20 °C cm−1. Those experiments may seem a breakthrough, however, the authors did not address how such large gradients could have developed in nature when the heat loss by conductive cooling was so slow.

Perhaps we can generalise and replace the term "thermal gradient" by a gradient in supersaturation induced by fluid saturation? The effect would be the same: the segregation of an H2O dominated fluid phase would locally raise the liquidus temperature, supersaturating the melt in crystalline phases and triggering spinifex growth (cf. Grove et al. 1997). A good example are the micro-scale spinifex pyroxenes in the solute spherules (Fig. 3g). The process would not be strictly isothermal because segregation of a hydrous fluid phase extracts heat from the system. Note though that a thermal gradient as large as applied by Faure et al. (2006) would not develop. A prerequisite though is that many (if not most) spinifex textured komatiites were "wet" melts, that H2O contents were high enough (c.f. Barr et al. 2009) to trigger fluid saturation at emplacement pressure.

Indeed, there is ample evidence that some komatiites were hydrous (Arndt et al. 1998). Komatiites show geochemical affinities to boninites and can be associated spatially with boninitic melts (Cameron 1985; Parman et al. 2001; Grove and Parman 2004; Wilson et al. 2003). Silica rich komatiites crystallised orthoenstatite spinifex instead of the higher temperature polymorph protoenstatite, suggesting that liquidus temperatures were depressed, most likely by H2O (Barr et al. 2009). Clinopyroxene compositions are indicative of crystallization at elevated pH2O (Parman et al. 1997). Komatiites can be rich in vesicles and amygdales (Beresford et al. 2000; Dann 2000, 2001; Fiorentini et al. 2012). And finally, komatiites may carry interstitial hydromagmatic amphibole in cumulates and differentiated sills (Grove et al. 1999; Fiorentini et al. 2012). Shore and Fowler (1999), in their Fig. 2, documented cm-sized orbicules in groundmass wedged between spinifex sheaves but did not appreciate the potential significance of that image. Such textures may form when hydrous melts segregate a fluid phase (Ballhaus et al. 2015; Fonseca et al. 2021) and solidify as emulsions.

Unfortunately, no elevated primary H2O has yet been detected in komatiites but there are reasons for this. Glasses suitable for IR analysis are rarely preserved (Nisbet et al. 1987) and then only in micron-sized melt inclusions along with quench microlites that need be homogenised with the glass prior to meaningful IR analysis (Berry et al. 2008). Komatiites are unusually hot melts even by taking into account the liquidus depression by H2O (Grove et al. 1999; Barr et al. 2009), hence they are susceptible to outgassing. All Archaean and palaeo-Proterozoic komatiites experienced a metamorphic hydration overprint (Grove et al. 1999) at least at lower greenschist facies, easily eliminating signatures of primary H2O.