The lithosphere-asthenosphere boundary (LAB) is fundamental to plate tectonics, with the overriding, brittle lithosphere “floating” on a ductile, convecting asthenosphere, allowing for plate motion such as divergence at mid-ocean spreading centers6 (e.g., East Pacific Rise). The plethora of geophysical and geologic techniques used to probe this boundary, has resulted in several complementary definitions. In a simplistic view, this boundary separates two distinct thermal regimes: conductive heat transport above the LAB versus convective heat transport below. The LAB can also be viewed as a boundary from a mechanical, rheological, or compositional perspective7. Taken together, these definitions help to outline, beneath mid-ocean spreading centers, a tent-shaped boundary that focuses melt rising from deep within the mantle towards the crust, leading to eventual eruption at the seafloor or accretion of gabbroic rocks that form the mid- to lower-crust8. Seismic and electro-magnetic probing across the LAB reveal slower velocities and higher electrical conductivities suggestive of a warmer and weaker regime beneath this boundary that is attributed to the presence of melt9,10. In addition, geodynamic modelling suggests melt extraction11 provides a mechanism to pool melt at the base of the LAB that is consistent with seismic and electro-magnetic observations near oceanic spreading centres12,13, including beneath the Juan de Fuca plate east of Axial Volcano14. The nature of this boundary in the crust and uppermost mantle in the vicinity of spreading centres, however, is uncertain, except for the presence of a narrow, 1-2 km wide mid-crustal axial magma lens (AML) that is generally observed at intermediate- to fast-spreading centres and separates the brittle sheeted dikes and pillow basalts from the lower mush zone, including melt-rich sills stacked between the AML and Moho15. Although tomographic images provide evidence of a wider halo of lower velocity material seated just beneath the AML4,16, extending 3-4 km on the ridge flanks, it has not been possible thus far to image it as a distinct boundary as observed for older lithosphere1. The ability to place geometrical constraints on the LAB deep into the crust beneath mid-ocean ridges would shed light on melt delivery into the lower- to mid-crust, which is critical to understand both eruptive and hydrothermal processes, and emplacement of the gabbroic section.
To investigate the dynamics of magma delivery beneath Axial Volcano, a three-dimensional (3-D) seismic reflection survey spanning 40 km by 16.3 km was collected aboard the R/V Marcus Langseth in 2019 at Axial Volcano, which sits at the intersection of the Juan de Fuca Ridge and Cobb-Eickelberg Hotspot (Fig. 1)17. Axial Volcano hosts several hydrothermal fields18, and has been the site of three eruptions19,20 over the past thirty years. Axial Volcano lies at a nearly constant water depth of about 1.4 km, is nearly flat and hosts a horseshoe-shaped 8 km x 3 km caldera that separates north and south rift zones of the Axial segment along the Juan de Fuca Ridge. An earlier two-dimensional (2-D) seismic reflection survey was collected at Axial Volcano in 2002 and highlighted the potential for complex interactions between adjacent magma reservoirs21,22,23,24,25, but lacked sufficient data density to link with origins of eruptions and with hydrothermal field locations. Three-dimensional seismic reflection data are markedly superior to 2-D in terms of accuracy of the final images and provide irreplaceable constraints on the spatial extent, continuity, orientation and dip of subsurface features, as demonstrated by prior 3-D surveys along the East Pacific Rise centered at 9°03’N26 and 9°50’N27, and a similar survey at Lucky Strike volcano on the Mid-Atlantic Ridge at 37°N28. Only through a comprehensive 3-D seismic reflection survey at Axial Volcano is it possible to image the complex pattern of AML reflections, underlain in places by vertically stacked magma sills in the lower crust15,22, to address the question of how and at what level melt is focused into and through the crust and their relationships with hydrothermal circulation and dike initiation.
This new 3-D seismic reflection volume reveals the presence of large-scale, funnel-shaped structures extending deep into the lower crust (2.0 s two-way travel time or ~5-6 km below the seafloor (bsf), Fig. 2) that appear to be the crustal signature of the elusive LAB at young ages; these reflections arise from a seismic impedance contrast across a distinct boundary, resulting in the illumination of this wide-spread interface. With the ability to map the downward extensions of this “expanded” AML deep into the lower crust (Figs. 1, 2a & 2b, 3; Supplemental Figs. 1, 2a & 2b, 3a & 3b), the distinction between the AML and LAB begins to fade as they are effectively one in the same structure (AML|LAB); melt trapped along this interface, driven by buoyancy along a thermal boundary, attempts to migrate toward the apex of this structure, where it is focused and primed for eruption, or fails and solidifies to become the emplaced gabbroic section.
Interestingly, the AML|LAB is funnel-shaped, not flat, and does not mimic the seafloor morphology beneath Axial Volcano and its environs. These funnel-shaped features (Figs. 2a & 2b, 3; Supplemental Fig. 1) constitute geometrical arrangements of melt that have never been seen before. This contrasts with an AML that is typically reported as flat, and whose depth is spreading-rate dependent and controlled by the maximum depth of hydrothermal circulation based on both prior observations and theoretical studies29. These funnel-shaped structures are imaged at four locations beneath Axial Volcano, out of which two are northwest (Funnel A) and southeast (Funnel B) of the horseshoe-shaped caldera; these features are not aligned with the rift arms but are rotated counter-clockwise approximately 25° and 35° to these features, respectively (Fig. 1). A third funnel-shaped structure (Funnel C) lies beneath the shallow bathymetric plateau east-southeast of the caldera, where it is abruptly displaced approximately 3.5 kilometers to the northeast from the downdip termination of Funnel B (Fig. 3, Supplemental Figs. 1, 2a & b). Lastly, a fourth funnel-like structure is seen beneath the New Dymond hydrothermal field and appears to overlap slightly with Funnel C, suggesting connectivity like that linking Funnels B & C (Fig. 3; Supplemental Figs. 1, 2a & 2b). Two of the four funnel-shaped structures (Funnel A & B) are positioned to directly focus magmas into the caldera (Fig. 2), while the third (Funnel C) and fourth (Funnel D) funnel-shaped structures terminate approximately 8 and 13 kilometers east-southeast of the caldera edge, respectively (Figs. 1 & 3). The magmas underpinning Funnels C & D are likely hotspot-sourced as the footprint of the Cobb-Eickelberg hotspot is positioned in this region southeast of Axial Volcano as inferred from a slight increase in crustal thickness determined using wide-angle data30.
Vertical slices perpendicular to the major axis of each of the funnel-shaped structures, or a combined transect along the apex of Funnels A & B, clearly show an elongated-dome appearance to the AML|LAB, spanning about 5 km and 22.5 km, respectively, with dips ranging from near flat to 50° (Figs. 2a & b). The morphology of the AML|LAB interface is in contrast to the vertically-stacked magma sills reported using conventional 2-D profiles15,22. Instead, we find that the funnel-shaped AML|LAB bounds these underlying sills and, in many cases, truncates them (Figs. 2a & b; Supplemental Figs. 2a & 2b, 3a & 3b). The fact that the underlying sills abut this reflector provides strong evidence that the bounding structure is the thermally-controlled LAB. The lateral dimensions of these stacked sills are at most 3.5 kilometers, and in some cases, much smaller. The shallowest section of the AML|LAB is located just southeast of the caldera and is expressed as a narrow ribbon of magma that shoals in two separate areas separated by 2.1 km, both of which sit at the updip termination of Funnel B (Fig.4; Supplemental Figs. 4 & 5). These locations are also coincident with the highest concentration of hydrothermal vents that comprise the International District field18. Furthermore, this ribbon-shaped melt body seems to underlie the initiation point of all three documented eruptions in 1998, 2011 and 201519,20.
The strength of the AML|LAB reflection depends primarily on the impedance (velocity*density) contrast across this boundary, although tuning within the AML can also cause variations in reflector strength31; other contributing factors to amplitude variation include attenuation and topographic focusing effects. To explore a map of reflection strength (a first-order filter to separate out regions of high melt concentration), we have employed a voxel visualization strategy whereby each seismic amplitude sample is represented as an elemental parallelepiped (a voxel in 3-D being equivalent to a pixel in 2-D), and only those datapoints with the highest seismic amplitudes values are visualised26. In this way, regions beneath Axial Volcano with the strongest reflections, and by proxy melt availability, can be viewed in three-dimensions. Projected in map view, the strongest reflections underlie the hydrothermal fields that are located along the northeast edge of the caldera (Fig. 4). Strong reflections are also seen near the intersection between the north rift zone and the caldera, with the largest footprint, however, observed southeast of the caldera beneath the extended plateau (Fig. 4; Supplemental Figs. 4 & 5). Strong reflections are also coincident with the perched AML farther to the southeast. In contrast, the distribution of strong reflections directly beneath the caldera is spotty and less continuous relative to other regions already highlighted. The two shallowest reaches of the AML|LAB (0.56 s and 0.65 s two-way travel time or an ~1.1 km bsf minimum depth23) separated by 2.1 km, can be observed using voxel visualisation (Supplemental Fig. 4), which as previously mentioned is near the initiation point for the recent eruptions. This narrow ribbon of high melt-fraction is coincident with PmeltS reflections (impinging on the AML as a P-wave, where it gets converted to an S-wave, then converted back to a P-wave at the seafloor) that were observed from a 2-D profile collected in 200221; strong PmeltS reflections are observed where the AML consists of nearly pure melt31.
Taking these new observations together, the simplest explanation for the large-scale funnel-shaped structures observed is that of a magma assimilation front, along which rocks from the overlying carapace are being remelted (Fig. 5); this is especially obvious at Funnels C and D, where magma has melted through the overlying lithosphere, resulting in perched magma bodies that are located a considerable distance from either Axial Volcano or its associated rift arms (Supplemental Figs. 2a & 2b). Therefore, the AML|LAB imaged here likely represents an assimilation front controlled by the thermal structure at Axial Volcano, in a configuration like that of melt underpinning the LAB for older oceanic lithosphere1.
Southeast of the caldera, the downdip termination of Funnel B is coincident but laterally offset 3.5 km from the updip termination of Funnel C. If viewed from a perpendicular slice through the major-axes of Funnels B & C at this offset, the dome-shaped AML|LAB structures “connect” at depth, suggesting that magmas beneath Funnel C could “leak” into Funnel B. This abrupt transition (Fig. 3; Supplemental Fig. 1) may owe its existence to magma migration along a failure-related fracture/surface in a manner analogous to Kīlauea Volcano on the south flank of Hawai’i32. Magma at Kīlauea Volcano is transported tens of kilometers along the east and southwest rift zones, which are associated with gravity-driven failure surfaces along the southern flank of Hawai’i; a similar process may be at play during eruptions at Axial Volcano. A failure surface would not only explain this abrupt offset but clarify how hot spot-related magmas are transported from Funnel C, into Funnel B, and ultimately to the caldera. It is unlikely, however, that the downdip termination of the LAB at Funnel B is that abrupt; instead, magmas are escaping from Funnel C, across a failure surface, then coalesce beneath the LAB associated with Funnel B, and thereby allow the updip extension to be seismically imaged. The downdip extension of Funnel B likely exists but is difficult to image due to a lack of sufficient melt presently trapped beneath this boundary. The ability to detect this boundary, which is related to the accumulation of magmas beneath it, will be greater during an episode of increasing magma supply from the mantle; such an episode in turn drives an active phase of assimilation as magma melts through the overlying brittle rocks.
A similar AML morphology is observed at a much smaller scale at three locations along the southern and northern East Pacific Rise at 17°20’S26,30 and 8°52’N31, and the Cleft segment of the Juan de Fuca Ridge at 44°46’N15, respectively, where the overlying sheeted-dike complex appears to be thinned by at least ~400 m over a lateral distance of ~5 km; poor seismic coverage at these locations, however, provides uncertainty to the origin and extent of these anomalies. Extended or wide AML reflections are also seen just north of the 9°03’N overlapping spreading center (OSC) on the East Pacific Rise26,35. A decoupling in melt supply between the northern and southern limbs of the OSC may allow magmas to accumulate beneath the LAB, producing anomalous AMLs that are imaged at ~4 kilometers in cross-axis width. A similar geometry may help explain the AML|LAB imaged near the CoAxial segment within this dataset (Fig. 1), where an extended AML|LAB reflection pointing updip toward the northern rift arm of Axial Volcano is imaged.
As the AML|LAB boundary plunges deep into the lower crust, active phases of magma assimilation would necessarily involve remelting of both the sheeted dikes (above the AML) and gabbroic rocks (above the downdip AML “extensions”) back into the reservoir (Fig. 5). The existence of magma assimilation at the gabbro-dike boundary has been well documented5,36, but there have been limited observations of the assimilation of gabbroic rocks in the lower crust37. At IODP hole 1256D, the lava sequence is 650 m thick, whereas the foreshortened dike sequence is only 350 m thick36, which could be due to the assimilation of the sheeted dikes at this location. Near the caldera at Axial Volcano, the average velocity structure of the upper crust is characterized by an ~600-800 m thick Layer 2A, underpinned by localized shoaling of the AML|LAB depth to ~1100 m bsf23, which suggests a thinned Layer 2B of ~300-500 m thickness; this observation mimics what was observed at IODP hole 1256D, providing additional evidence for significant assimilation along this interface beneath Axial Volcano. The estimated dips of the AML|LAB interface at Axial Volcano range from nearly flat up to 50° that mimic the measurements of magmatic layering and foliation seen in upper gabbroic sequence within the Oman ophiolite with dips that range from 0° to ~40° giving an average dip of 18-19°38; melt migration and assimilation along the AML|LAB interface may provide a mechanism to impart these observed dips that are subsequently locked-in as the thermal boundary migrates due to plate spreading.
Analogue studies related to inflation-deflation processes within volcanic systems39 would predict that the shallowest reaches of the magma chamber would be the most favorable for dike initiation followed by emplacement, which may match the recent eruption history at Axial Volcano. In fact, the southward and northward propagations of dikes during the 1998 and 2015, and 2011 eruptions, respectively, seem to be associated with the two peaks on the ribbon-shaped melt body (Fig. 4 and Supplemental Fig. 4), indicating that the shallowest anatomy of the magma bodies may control the eruption initiation and the direction of dike propagation. Interestingly, dike propagations occur along the Axial rift zones, not where the most expansive melt-rich bodies are observed in the sub-surface, indicating that dike propagations, and hence eruptions, occur in regions where extensional stress is maximum (Fig. 4). The limited observation of strong AML|LAB reflections beneath the north and south rift zones24 further confirms the above observations. Another interesting question is whether the extent of high seismic reflection amplitudes southeast of the caldera, likely associated with melt enhancement, increased during the eight intervening years between the 2011 eruption and data acquisition; given the proximity of this feature to the south rift zone, it may indicate that the next eruption sequence could propagate to the south. On the other hand, the narrow melt ribbon imaged in 2019 near the northeast caldera wall may have been diminished in size as deeper-seated magmatism transited through this region during the 2015 eruption. As the eruptions appear to be initiating from the shallowest parts of the magma chamber, and are responsible for the formation of Layer 2, the lower crust is formed below the AML|LAB boundary, and hence there could be a decoupling between the lower and upper crustal formation as well. This might also help explain the rhomboid shape of Axial Volcano.
As assimilated rocks contain hydrous minerals formed due to the deep-reaching hydrothermal circulation, associated reservoir contamination would cause changes in the chemistries of erupted lavas towards more plagioclase-rich trends40,41, and to an extreme, silica-rich magmas are possible as observed in Iceland42,43. The evolved lavas observed at Axial Volcano44 and at many mid-ocean spreading centers5 were thought to involve petrologic differentiation through isolation as potentially provided by sub-AML|LAB sills45. Our new findings, however, would suggest that two mechanisms simultaneously exist beneath Axial Volcano (Fig. 5): (1) stacked vertical sill structure may provide pathways that promote isolation of magmas giving rise to greater differentiation, and (2) through assimilation of the overlying hydrated substrate, an independent mechanism exists to drive differentiation towards more evolved magmas. At Axial Volcano, these two mechanisms exist together and could potentially work in tandem to enforce trends towards further differentiation to help explain the petrologically differentiated lavas (Group 1 with low Mg #: <7.9) observed at this site44. The more mafic lavas such as those observed during the first part of the 2015 eruption (Group 2, high Mg # > 7.9) are likely not sourced from magma that resides along the AML|LAB interface, but rather from a deeper source that mostly bypasses both contaminated melts and underlying sills, and is directly emplaced into dike feeder system46,47. Petrological studies of historical Axial lavas flows predominantly suggest an evolved or Group 1 origin; in contrast, the last eruption in 2015, and a time span from 1200-1400 CE, contained less evolved Group 2 lavas. Together, these observations suggest that Axial Volcano has been constructed more often by evolved lavas that have either migrated along the AML|LAB boundary and/or isolated in sub-AML|LAB sills, with less frequent contributions by primitive lavas that bypass mid- to lower-crustal processes and directly erupt on to the seafloor (Fig. 5). A subset of 2015 lavas along the northern rift zone that are more petrologically evolved may have tapped into the contaminated melt zone along the AML|LAB interface near the intersection of the northern rift zone and caldera (Fig. 4).
The ability to image the wide-spread AML|LAB interface has almost certainly benefited from a strong magmatic phase currently underway at Axial Volcano44,46,47. The lack of significant topography relative to most mid-ocean spreading centre environments also provides near ideal conditions for seismic imaging at depth. These factors have helped enable a new view into the dynamics of melt delivery near the base of the crust and redistribution not only along the AML|LAB toward the caldera, but also along failure surfaces that appear to exist within this volcano as the edifice succumbs to a combination of body forces and magmatism. These images also lend support to a mechanism of assimilation of hydrated rocks, which can impart significant influences on the chemistry of lavas erupted at Axial Volcano, and more generally, elsewhere. When the current phase of robust magmatism finally wanes at Axial Volcano, so will the degree of assimilation, and solidification/accretion along this boundary will take place as the size and scope of the AML|LAB reduces accordingly. Within this magma-restricted environment, one might expect less petrologic diversity as both contamination through assimilation and multi-sill-based isolation becomes less common, and so does the prevalence of evolved lavas. In turn, the absence of a robust magma supply makes imaging the downdip extensions of this boundary more difficult, resulting in a shrinking AML|LAB beneath Axial Volcano that becomes indistinguishable from smaller AMLs seen beneath other spreading centers. Thus, these episodic upswings or super-cycles of magma delivery into the crust, should not only help to reveal the downdip extent of the LAB, but should also leave, within erupted lavas, a cyclical record of contamination through assimilation of erupted lavas41 that may explain the bi-modal distribution of lavas erupted at Axial Volcano44.
Our results have implications for other large volcanic provinces formed by ridge-plume interaction, such as Iceland. In Iceland, a thick crust (30-40 km)48 is formed by the interaction between the Iceland Plume and the Mid-Atlantic Ridge. As it is difficult and expensive to seismically image magma systems on land, the nature of the crustal AML|LAB beneath Iceland is poorly understood. Nevertheless, a record of extensive assimilation has been observed in the petrology of erupted lavas42,43. We suggest that sills are likely not emplaced as a series of vertically stacked sills beneath the active rift zone, but that they are likely distributed in three-dimensions, and present near the AML|LAB boundary, where assimilation occurs. A similar process may also occur at slow and ultra-slow spreading ridges, where the AML|LAB boundary is likely to be three-dimensional and ephemeral, and where melt would migrate along the AML|LAB boundary towards the segment centre, forming the central volcano, such as the Lucky Strike volcano28 in the Atlantic and the 50°28’E volcano on the Southwest Indian Ridge 49. The petrology of lavas from the above volcanic centres does indicate the existence of assimilation41, 50, implying that the crustal AML|LAB and magma assimilation might be prevalent at other volcanic centres.