T. D. Jones1, D. R. Davies1, I. H. Campbell1, G. Iaffaldano2, G. Yaxley1, S. C. Kramer3 and C. R. Wilson4,5
1Research School of Earth Sciences, The Australian National University, Canberra, Australia 2Department of Geosciences and Natural Resource Management, University of Copenhagen,
Denmark
3Department of Earth Science and Engineering, Imperial College, London, UK 4Lamont-Doherty Earth Observatory, Columbia University, New York, USA 5Department of Terrestrial Magnetism, Carnegie Institution of Washington, USA
Copyright c2017 by Springer Nature. Reproduced with permission. The official ci- tation that should be used in referencing this material is T. D. Jones, D. R. Davies, I. H. Campbell, G. Iaffaldano, G. Yaxley, S. C. Kramer, C. R. Wilson, 2017. The con- current emergence and causes of double volcanic hotspot tracks on the Pacific plate. Nature, 545, pp.472-476. The use of this information does not imply endorsement by the publisher.
Abstract
Mantle plumes are buoyant upwellings of hot rock that transport heat from Earth’s core to its surface, generating anomalous regions of volcanism that are not directly associated with plate tectonic processes. The classic and best-studied example is the Hawaiian-Emperor chain. However, the emergence of double-track volcanism along this chain Jackson et al. (1975) – namely the Loa and Kea tracks – and the sys- tematic geochemical differences between them Tatsumoto (1978); Abouchami et al. (2005) have remained enigmatic. Here we demonstrate that their emergence co- incides with the appearance of other double volcanic tracks on the Pacific plate and a recent azimuthal change in Pacific plate-motion. We propose a three-part model that explains the evolution of Hawaiian volcanism: (i) mantle flow beneath the rapidly moving Pacific plate strongly tilts the Hawaiian plume and leads to lat- eral separation between high and low pressure melt source regions; (ii) the recent azimuthal change in Pacific plate-motion exposes high and low pressure melt prod- ucts as geographically distinct volcanoes, explaining the simultaneous emergence of double-track volcanism across the Pacific; and (iii) secondary pyroxenite, formed as eclogite melt reacts with peridotite Yaxley and Green (1998), dominates the low pressure melt region beneath Loa-track volcanism, yielding the systematic geochem- ical differences observed between Loa- and Kea-type lavas Frey and Rhodes (1993); Norman and Garcia (1999); Abouchami et al. (2005); Weis et al. (2011); Jackson et al. (2012); Frey et al. (2016). Our results imply that the formation of double-track volcanism is transitory and can be used to identify and place temporal bounds on past plate-motion changes.
Introduction
The mantle plume hypothesis predicts that a single age-progressive chain of volcan- ism will form as a tectonic plate passes over a comparatively fixed plume conduit. However, for the past ∼3 Myr magmatism along the Hawaiian-Emperor chain has manifested as two sub-parallel volcanic tracks, each with a distinct geochemical sig- nature (Fig. 4.1). Double volcanic chains have been observed elsewhere, at Easter, Foundation, Galapagos, Marquesas, Samoa, Society and Tristan-Gough Huang et al.
Figure 4.1: Bathymetric map of recent Hawaiian volcanism, highlighting the Loa and Kea tracks: solid lines represent the geochemically distinct Loa (blue) and Kea (red) volcanic trends Abouchami et al. (2005); Weis et al. (2011). The dashed lines are an estimate of the future projection of these trends based upon our hypothesis. Prior to the emergence of the Loa and Kea tracks, Kea-type lavas are overlain by Loa-type lavas at Ko’olau and Kaua’i Garcia et al. (2010); Weis et al. (2011). The encircled number indicates the approximate age at which Loa trend lavas first appeared at Ko’olau.
(2011); Payne et al. (2013); Weis et al. (2011); Harpp et al. (2014). Four of these double-tracks occur on the Pacific plate and emerge within the past ∼2-4 Myr, concurrent with the appearance of the Loa and Kea trends at Hawaii (see Supple- mentary Material to Paper 3: Fig. 7.1). Indeed, all volcanic chains on the Pacific plate that have displayed persistent volcanism for the past 5 MyrGripp and Gor- don (2002); Clouard and Bonneville (2005) show this trend, with the exception of the McDonald chain, which lacks the clear age-progressive characteristic often as- sociated with mantle plumes. Although double-track volcanism has been identified across the globe, Hawaii has received the most attention Tatsumoto (1978); Frey and Rhodes (1993); Norman and Garcia (1999); Abouchami et al. (2005); Bianco et al. (2005); Garcia et al. (2010); Weis et al. (2011); Ballmer et al. (2011); Jackson et al. (2012); Hofmann and Farnetani (2013); Garcia et al. (2015); Frey et al. (2016). Compositional differences among Loa and Kea track volcanics are well documented Tatsumoto (1978); Frey and Rhodes (1993); Norman and Garcia (1999); Abouchami et al. (2005); Weis et al. (2011); Jackson et al. (2012); Frey et al. (2016). Loa trend basalts typically have lower 206Pb/204Pb, 143Nd/144Nd, 176Hf/177Hf, Ti/Na
and CaO, and higher 187Os/188Os, 87Sr/86Sr, SiO2 and Ni, than those of the Kea trend Abouchami et al. (2005); Weis et al. (2011); Jackson et al. (2012). Many of these differences can be explained if the source region of Loa trend basalts contains higher proportions of eclogite and/or pyroxenite relative to the Kea trend Hauri (1996); Sobolev et al. (2005); Jackson et al. (2012), fuelling debate on the Hawaiian plume’s thermo-chemical structure. Hofmann and Farnetani Hofmann and Farnetani (2013) recently outlined the two prevailing views: (i) the Hawaiian plume conduit is bilaterally zoned, reflecting large-scale compositional heterogeneity at the base of the mantle that has been transported into, and preserved within, the plume during its ascent towards Earth’s surface Weis et al. (2011); and (ii) the Hawaiian plume is a mixture of enriched recycled material within a relatively depleted peridotite matrix, whereby shallow processes, such as differences in lithospheric thickness Ballmer et al. (2011) or spatial variations in the plume’s thermal structureBianco et al. (2005), lead to compositionally distinct melting trends. A weakness of the first view is that the density contrasts predicted for eclogite and/or pyroxenite will prevent bilaterally zoned plumes from developing Jones et al. (2016). Moreover, neither view explains the emergence of these features as geographically distinct volcanic chains at ∼3 Ma Jackson et al. (1975).
In one of few studies addressing this issue, Hieronymus and Bercovici Hieronymus and Bercovici (1999) demonstrate that double volcanic chains can develop with the introduction of an additional off-axis volcano following a change in plate-motion. However, their study focusses on lithospheric properties and does not explicitly consider the plume underlying Hawaii or address the geochemical characteristics of the two volcanic chains, thus precluding an understanding of the source and melting conditions controlling the geochemistry of lavas across this region. Here we account for these features in a three-part model that provides a dynamical plate scale explanation for the concurrent emergence of double-track volcanism across the Pacific plate and the origin of their systematic geochemical differences.
Laboratory and numerical experiments demonstrate that mantle plumes will be de- flected by mantle flow, particularly under fast moving plates like the Pacific, with tilting often towards the direction of plate-motion Griffiths and Richards (1989). At Hawaii, such a deflection is borne out by seismic imaging, with recent regional Cheng et al. (2015) and global French and Romanowicz (2015) tomography mod-
Figure 4.2: Schematic diagram of the tilted Hawaiian plume, the overlying Pacific plate and associated surface volcanism: a, cross-section prior to∼3 Ma, where the plume is deflected in the direction of plate-motion, with surface volcanism characterised by deep Kea-type lavas overlain by shallow Loa-type lavas as the plate migrates over the melt zone. b, cross-section immediately after a plate-motion change, with plate-motion now oriented at an oblique angle to the direction of plume tilting and the long axis of the melt region, exposing the melt source region to a greater area of lithosphere and producing the two parallel Kea and Loa trends, which sample from shallow and deep portions of the melt zone respectively. c,d, top views of (a) and (b) highlighting the change in geometry between Loa and Kea trend lavas following the change in plate-motion.
els showing a slow shear-wave velocity anomaly that has a tilt of ∼45◦ orientated
towards WNW in the upper mantle. As illustrated in Fig. 4.2a, plume tilting in the direction of plate-motion causes lavas at the base of successive shield-volcanoes to be derived from deep melting and later be overprinted with the lavas produced through shallower melting. At Hawaii, from ∼ 5−3 Ma, both Loa and Kea-type lavas are found within individual volcanoes, such as Ko’olau Fekiacova et al. (2007) and Kaua’i Garcia et al. (2010); Weis et al. (2011), with deep drilling revealing that Ko’olau evolves from Kea-type lavas at its base to Loa-type lavas above. Loa- type lavas are also found offshore Oah’u, with both shield and tholeiitic post-shield volcanism at Kaua’i having a Loa component Garcia et al. (2010). We therefore propose that the Kea-type lavas at Kaua’i and Ko’olau were derived from high pres- sure melts, with Loa-type lavas generated at lower pressures (Fig. 4.2a and Fig. 4.2c). This is consistent with studies that suggest Loa-type magmas have under- gone a higher degree of partial melting than their Kea-type counterparts Norman and Garcia (1999); Hofmann and Farnetani (2013).
Results and Discussion
The appearance of Loa and Kea, as geographically distinct volcanic tracks, ap- proximately coincides with a bend in the volcanic island chain (Fig. 4.1), indicat- ing a change in the relative motion between the Hawaiian plume and the Pacific plate. Although an azimuthal change in Pacific plate-motion is evident from high- resolution multi-beam bathymetric data along the Pitman Fracture Zone Auster- mann et al. (2011), it remains difficult to isolate in the record of stage Euler vectors from the recent and higher-resolution finite rotations of Wessel and Kroenke Wessel and Kroenke (2008). This is likely due to the increased impact of noise on the in- ferred kinematics Iaffaldano et al. (2012), which manifests as erratic changes in the Pacific Euler pole’s location since the Early Neogene (Fig. 4.3a). In Fig. 4.3b we illustrate the path of the Pacific Euler pole after mitigating the impact of data noise Iaffaldano et al. (2012) (see Supplementary Material to Paper 3, section 7.2.1). Our results show that over the past∼10 Myr the Pacific pole has wandered northwards, causing a clockwise rotation and a progressively larger northwards component in the direction of plate-motion at Hawaii. The most rapid polar migration occurred over the past 4.2 Myr (Fig. 4.3d), coinciding with the appearance of double volcanic tracks across the Pacific plate and corroborating the prediction made by Hierony- mus and Bercovici Hieronymus and Bercovici (1999). As illustrated in Fig. 4.2b, when the plate changes direction its motion will initially be oriented at an oblique angle to the direction of plume tilting. As a consequence, shallow and deep melts rising vertically will interact with different regions of lithosphere, thus allowing them to erupt through geographically distinct volcanic edifices at the surface, which we predict constitute the Loa and Kea tracks, respectively (Fig. 4.2b and Fig. 4.2d). As the plume responds to the modified flow regime, it will eventually realign to the new direction of plate-motion Griffiths and Richards (1989), causing the lava types to overprint once again.
Our hypothesis requires basalts of the Kea track to form at higher pressure than those of the Loa track, which will lead to differences in their erupted composition. However, key geochemical differences between Kea and Loa trend basalts cannot be reconciled by melting a homogeneous mantle at different pressures. As noted above, it has been proposed that these differences occur because the source region
Euler -pole: loca tion Euler -pole: r at e of motion (a) (b) (d) (c) Noise-mitigated Original -
Figure 4.3: Evidence for a recent change in Pacific plate-motion: a, location of the Euler pole (in red) for motion of the Pacific plate (PA, in green) relative to the hotspot reference frame, since∼24 Ma, as reconstructed by Wessel and Kroenke (2008), with Euler pole symbols becoming larger for stages closer to the present-day. The Euler pole, about which the Pacific plate rotates with respect to the hotspot reference frame, is currently located to the south of Australia, offshore Antarctica and, hence, at Hawaii, Pacific plate-motion is directed towards the northwest;b, as in a, but following noise-mitigation Iaffaldano et al. (2012) (see Supplementary Material to Paper 3, section 7.2.1); c/d, the rate at which the stage Euler poles in panels a/b wandered, respectively. Bounds on the time intervals correspond to the middle of stages from the reconstruction of Wessel and Kroenke (2008), whilst thin lines show the confidence ranges inferred from 1 million samples of Euler poles, drawn using the associated covariances. Solid and dashed lines show the intervals where the most-likely 10 and 20% of samples fall, respectively.
of Loa track basalts contains a higher proportion of pyroxenite relative to the Kea track Hauri (1996); Sobolev et al. (2005); Jackson et al. (2012). For our model to be consistent with this interpretation, the pyroxenite component has to have a higher solidus temperature than the peridotite component, thereby allowing it to dominate the low pressure source region that feeds Loa track volcanism.
Laboratory experiments, using natural compositions with synthetic eclogite and peridotite mixtures, show that this could be the case at pressures above ∼ 2 GPa Yaxley and Green (1998) (Fig. 4.4a); note that 2 GPa is the pressure at which the thermal divide, described below, becomes ineffective Lambart et al. (2013). Over the likely range of plume melting pressures, the eclogite solidus temperature is 100-200 K below that of peridotite (Fig. 4.4b). As a consequence, the first component to melt in a heterogeneous plume will be eclogite. As the plume continues its ascent, the composition of eclogite melts will move progressively along the clinopyroxene- garnet cotectic towards the thermal divide that separates the eclogite and peridotite liquidus fields Yaxley and Green (1998); Herzberg (2011) (Fig. 4.4a). These silica saturated melts will be out of equilibrium with olivine and orthopyroxene and will consequently react with the adjacent peridotite to produce secondary garnet-bearing aluminous pyroxenite that lies on the thermal divide Yaxley and Green (1998). Melting on the peridotite side of the divide starts at a lower pressure, at the four- phase olivine-orthopyroxene-clinopyroxene-garnet pseudo eutectic (Fig. 4.4a and 4.4b), to produce silica under-saturated melt, which can escape from the mantle because it is in equilibrium with the surrounding peridotite Yaxley and Green (1998). Melting of the secondary pyroxenite starts at even lower pressures, at the two-phase cotectic on the thermal divide, and has a higher solidus temperature than both the eclogite and peridotite four-phase pseudo eutectics (Fig. 4.4a) Yaxley and Green (1998); Herzberg (2011). We emphasise that the solidi of both peridotites and pyroxenites are sensitive to their composition and that existing experimental studies on pyroxenite melting, with the exception of Yaxley and Green Yaxley and Green (1998), were carried out on samples that lie off, and in some cases well away from, the thermal divide between the eclogite and peridotite eutectics. These experiments are not directly relevant to the scenario proposed here.
We next use 3-D dynamic modelling of an upwelling mantle plume to quantitatively validate our hypothesis (see Supplementary Material to Paper 3, section 7.2.2 and
Figure 4.4: Phase relations and solidus curves: a, Schematic liquidus phase relations for the peridotite–basaltic system at 3.5 GPa, modified from Yaxley and Green (1998). The green circle labelled ‘P’ represents average peridotite mantle composition, the pink circle labelled ‘E’ represents average eclogite composition, and ‘C’ is the cotectic on the thermal divide at which secondary pyroxenites begin melting. The liquidus fields are labelled Ol = olivine, Ga = garnet, Opx-Cpx = pyroxene and Co = coesite. Blue lines represent melt pathways with black arrows indicating increasing temperature (which is considered analogous to decreasing Pressure). The green, pink and yellow regions are psuedo-eutectic melt compositions for peridotite, eclogite and pyroxenite, respectively. Note that the arrows on the melt pathways point up temperature to ‘C’, showing that the two-phase cotectic on the thermal divide has a higher solidus temperature than both four phase eutectics. Therefore, pyroxenites forming on the thermal divide will have a higher solidus temperature than the peridotite eutectic and therefore melt at lower pressures Yaxley and Green (1998). b, Pressure temperature conditions for peridotite (green) and eclogite (pink) melting from Katz et al. Katz et al. (2003) and Yasuda et. al Yasuda et al. (1994), respectively. The pyroxenite solidus (yellow line) is inferred to be 50 K above the peridotite solidus (see text). The solid red line highlights the plume temperature from the model presented in Fig. 4.5, with the kink marking the depth at which the uppermost part of the plume approaches the base of the lithosphere.
7.2.3). In our simulation, the plume rises vertically through the higher viscosity lower mantle but tilts in the lower viscosity upper mantle due to simple shearing induced by the overlying plate (Fig. 4.5). Melt regions are mapped out in terms of the melt fraction (see Supplementary Material to Paper 3: Fig. 7.2) and the melting rate (Fig. 4.5), using parameterizations for peridotite, eclogite and pyrox- enite (see Supplementary Material to Paper 3, section 7.2.4). As is clear from Fig. 4.5, the zone of maximum melt productivity for each component is vertically offset, which is a direct consequence of the distinct solidus and liquidus relationships of the individual components (Fig. 4.4). The horizontal separation between regions of maximum melt productivity, however, is principally induced by the plume’s tilt. The predicted tilt, which closely matches that imaged for the Hawaiian plume, leads to a horizontal offset of∼40 km between the regions of maximum melt productivity for peridotite and pyroxenite, similar to the observed distance between Loa and Kea track volcanoes.
Figure 4.5: Simulation of a 3D mantle plume beneath a moving plate: the plume is contoured at a temperature of 1673 K. The model domain is outlined by thick black lines. plate-motion is indicated by white arrows. Melting of eclogite, peridotite and pyroxenite is contoured at half the maximum melt rate, thus highlighting the regions of maximum melt productivity for each component (see Supplementary Material to Paper 3, section 7.2.4). For eclogite, melting initiates at a depth of∼
240 km (8 GPa) and reaches a maximum at∼200 km (6.6 GPa), whereas peridotite and pyroxenite melting initiate at∼180 km (5.8 GPa) and ∼160 km (5.3 GPa), reaching a maximum at∼155 km (5.1 GPa) and∼135 km (4.5 GPa), respectively. The blue and red cylinders illustrate the melt extraction pathways for Loa and Kea source regions, respectively, under the assumption of vertical melt ascent. a, perspective view of the plume where plate-motion is parallel to the direction of plume tilting and surface volcanism evolves from magmas that form at high pressure, to magmas that form at lower pressure. b, as in (a) but viewed from behind the plume. c, perspective view of plume 2.5 Myr after the imposed, instantaneous 20◦plate-motion change, allowing double-track volcanism to emerge at the surface and sample distinct portions of the melt zone. The plume has also undergone a degree of vertical rebound, reducing its tilt and the associated separation between peridotite and pyroxenite melting. d, as in (c) but viewed from behind.
instantaneous 20◦ change in the prescribed plate-motion direction is imposed (Fig. 4.5c and 4.5d). The plate initially moves at an oblique angle to the direction of plume tilt but, over time, the underlying plume adjusts and realigns to the new direction of plate-motion (Fig. 4.6a). A major consequence of the transient offset between these azimuths is that high pressure melts are prevented from overprinting low pressure melts, thus allowing them to erupt along geographically distinct volcanic chains. In the context of Hawaii, this requires volcanism on the Loa (south-western) side of the volcanic track to form by shallow melting of peridotite and pyroxenite and