Our samples were dredged by RV Thompson Cruise TN365 on the northeastern Marion Rise in March 2019 on the eastern third of the ~255-km long amagmatic supersegment between the Discovery and Indomed FZ’s. There two orthogonal to spreading magmatic ridge segments are linked by an intervening 67-km long oblique-amagmatic segment oriented at ~15° to the spreading direction (Fig. S1). The entire region is an area of extreme low melt productivity, where only scattered volcanics lie on serpentinized mantle between widely spaced magmatic segments. The explosive lavas were recovered in Dr 30 at 40°08.6’S, 45°28.2’E high on the southern rift valley wall of the 70-km long 45°30’E ridge segment (Fig. 1). The dredge was from 2,195 to ~1,850-m depth up the north side of a 1-km wide cratered debris cone, with a secondary ~200-m wide cone at its center. Additional candidate craters occur on the rift valley walls and adjoining rift mountains (Fig. 1). While some are locally cut by normal faults, others are undisturbed indicating likely off-axis volcanism. Possible craters in the rift valley are obscured by fissure eruptions along the axial neovolcanic zone. The physiography indicates two overlapping volcanic phases – explosive and effusive, one occurring largely in the rift valley, the other manifesting over a broader zone into the rift mountains. Lavas were also recovered from the 45°30’E segment at 44°09.54’S, 45°14.73’E on the south wall of the inside-corner high at its intersection with the oblique amagmatic segment in Dr 31 (Fig. S1). Some 37-km to the southwest at 40°25.98’S, 44°59.82’E Dr. 27 recovered additional basalts from the north wall of the 20-km long 45°E magmatic segment (Fig. S1).
Dr 30 recovered fresh pillow basalt, glassy volcanic scoria and lava bombs ranging from 6 to 15 cm in size (Fig.’s 2a-c, S3a-b), and a large amount of siliceous sinter, some with thin hydrogenous manganese coverings (Fig. 2d). The lavas have 15 to 25 vol.% 2-3 mm plagioclase and ~1% <2 mm olivine phenocrysts and 22-61 vol.% vesicles measured by cross-sectional computed tomographic imaging and microscopy (Fig. 2a-b). Some of the glassy scoria are partly covered with very-thin silica sinter (Fig. 2c). SEM imaging and EDS show the sinters formed beneath a thin sediment layer with numerous silica balls pseudomorphing foraminifera, and abundant round holes left by their dissolution (Fig. S2).
Dredge 31 contained large chunks of ultraphyric diabase or basalt with 68% ~10-mm plagioclase and 2% 1-mm olivine phenocrysts. One vuggy weathered and altered pillow basalt with zeolites filling 1 to 3 mm coarse vesicles may represent a flow top where vesicles accumulated and grew. No glass is preserved.
Dr 27 dredged light to moderately weathered aphyric to plagioclase phyric pillow basalts, and moderate to heavily weathered avesicular ultraphyric basalts and diabase. The pillow basalts have 0-10% vesicles, (ave. 5%, n=25), and 0-10% plagioclase phenocrysts, several with 1-2% olivine phenocrysts. The contrast in vesicularity between Dr 27 and 30 glasses is striking, with much higher volatiles indicated in the latter (Fig. S3c-d).
Rock and Mineral Chemistry
Analytical methods, and isotopic, major and trace element data for the sinters, lavas, glasses, and minerals are in the supplemental materials (Tables S1 to S8). Dr 30 glasses range from olivine tholeiite to nepheline normative alkali basalt (0 to 3.67 wt.% Ne) similar to ocean island basalts with elevated La and K2O (0.93 to 1.28 wt.%), La/Yb, K/Ti, 87Sr/86Sr, and low 143Nd/144Nd (Fig.’s 4, S4c, S5, S6; Table S2). However, K/Ti vs Mg# and isotopic compositions are distinctly different from Marion and Crozet Island basalts (Fig. 4b, d). Though straddling the MacDonald-Katsura line between Hawaiian tholeiites and alkali basalts 5 (Fig. 4a), the majority lie in the alkali field. 3He/4He ratios in the Dr 30 and Dr 27 glasses (6.6 to 8.2 RA) (Table S7) are typical for Marion Island lavas and normal MORB 6 (Fig. S7), but well below those of Hawaiian and Icelandic plume basalts as well as Crozet Island Lavas 7. The Dr 30 heavy isotopes and K2O-TiO2 are also distinct from Marion or Crozet (Fig. 4b, d). This is consistent with a distance >1,000 km from Marion Island, and the intervening Discovery and Indomed FZ’s, which would block sub-axial mantle flow from either direction 8. The Dr 31 and 27 basalts are alkali-rich quartz normative tholeiites (1.12 to 8.45 wt.% qtz) lying below the MacDonald-Katsura line in the silica-alkalis plot (Fig. 4a), with flat to light REE depleted trace-element patterns (Fig.’s 4c, S5). The fractional crystallization trend between Mg# and other oxides defined by Dr 31 and 27 volcanic rocks is clearly different from that of Dr 30 glasses and lavas (Fig. S4).
The silicic sinters from Dr 30 are nearly pure amorphous SiO2 with little Fe oxides/hydroxides, except for parts entraining pelagic sediment. Oxygen isotopes indicate they precipitated from 26°C to 45°C (Table S8), but differ from other low-temperature precipitates at spreading centers (Juan de Fuca, Lau Basin and SWIR etc.) due to their low iron contents. They likely formed by venting of hydrothermal fluids due to shallow seawater circulation within the cooling lavas.
Nearly all the Dr. 30 basalt plagioclase and olivine phenocrysts are primitive xenocrysts (An80.4 - 85.9, An82.6 ave.; Fo81.5-89.4, Fo87.3 ave.) and in disequilibrium with the enclosing basalt (Fig. 3c; Table S4). Plagioclase crystals are rounded (Fig 3a) to sub-equant, indicating variable resorption by the host magma; though in several samples there is a second population of subhedral to euhedral plagioclase laths, which indicates one or more stages of magma mixing occurred during ascent to the seafloor 9-11.
Dr 30 olivine melt inclusions and plagioclase xenocrysts have 87Sr/86Sr compositions intermediate to the Dr 30 and Dr 27 glasses, with the inclusions lying close to the Dr 27 glasses and the plagioclase phenocrysts close to the Dr 30 glasses (Fig. 3b; Table S3). Moreover, they range from LREE enriched, like the enclosing alkali basalt, to flat N-MORB-type patterns (Fig.’s 4c). This indicates that the alkali basalt incorporated a large population of xenocrysts but little interstitial melt from a tholeiitic crystal mush, preceding and possibly triggering eruption, with partial requilibration of the xenocrysts and their inclusions. This is consistent with the Dr 30 spinel Cr/(Cr+Al), which ranges from 39.1 to 56.6 with TiO2 from 0.34 to 1.62 wt.% (0.99 ± 0.29 ave.) (Table S4). These are typical MORB spinels, where high-Al coexisting with high-Cr spinel likely reflects high pressure crystallization followed by shallow resorption and crystallization of new Cr-rich spinel during magma mixing 12, 13.
Volatiles in our basalts are highly variable, with low CO2 (85 ppm ave.) unlike volatile-rich MORB from the Juan de Fuca Ridge or the Mid Atlantic ridge (MAR) at 14° and 34°N (Fig. 5). Dr 30 olivine melt inclusions, however, have significantly higher CO2 (550 ppm ave.; Table S6), than in Dr 30 glasses or Dr 27 and 31 olivine melt inclusions. As CO2 does not correlate with the other volatiles, this would reflect preferential degassing during ascent (Fig. S8a, c). Due to explosive degassing, then, the Dr 30 volatiles do not tell us that much about primary melt concentrations, and may account for the low CO2 in the glass compared to the DR 27 glasses, and DR 27 and 31 melt inclusions where there is no evidence for explosive degassing.
Any assimilation of seawater (18,980 ppm Cl) during magma fountaining and fall back, or interaction with hydrothermally altered rocks, will drastically affect melt Cl concentrations. Highly vesiculated basalt fragments in the Dr 30 scoria provides such evidence (Fig. S3b). Chlorine is much higher in the Dr 30 (554 ppm ave.) than in DR 27 glass and melt inclusions (Fig. S8), and both glasses and melt inclusions have very low F/Cl indicating assimilation of seawater during eruption, which could contribute to high H2O contents in the Dr 30 glass and melt inclusions (5,254 and 5,248 ppm ave.). Successively higher F/Cl in the DR 27 glasses, and DR 27 and DR 31 melt inclusions, then, indicate decreasing interaction with seawater (Table S6). The lack of correlation between volatiles in the Dr 27 glass and inclusions indicates that their concentrations were controlled by stochastic mixing of variably fractionated tholeiitic melts, and unsurprisingly S, H2O, and F in the Dr 30 and Dr 27 olivine melt inclusions are quite similar.
H2O, S, Cl and F due show good correlations with each other in the Dr 30 glasses (Fig. S8a, b), and some weak correlations in the Dr 30 Ol-melt inclusions. Both the glasses and inclusions have high H2O (0.5 wt.% ave.) ranging up to 0.8 wt.% in inclusions (Table S5) compared to typical MORB (Fig. 5) (< 0.2% 14). Thus, the alkali basalt was saturated only in CO2 during ascent, while the remaining volatiles were largely enriched by crystal fractionation consistent with only a small amount of melt in the tholeiitic crystal mush. The scatter in the olivine melt inclusions indicates their concentrations were modified during entrainment, as were the trace element and isotopic compositions of the xenocrysts.
Including their vesicularity, the Dr 30 lavas and scoria had total CO2 at the depth of eruption ranging from 0.79 wt.% to 4.10 wt.%, assuming the vesicles contained only CO2. However, the explosive threshold for CO2 is ~7.8 wt.% if the generally accepted 75 vol% CO2 is needed to cause explosion 15. Thus, the volatiles in the melt reservoir approximated by the olivine melt inclusion contents are nominally insufficient for the volcanic explosion at the dredge depth 16, 17. Subsurface accumulation of exsolved volatiles, primarily CO2, as in strombolian or vulcanian eruptions 16, seem to be required to explain the bombs and scoria in Dr 30.