Asthenospheric zircon below Galápagos dates plume activity


 Mantle plumes are active for long periods of time1,2, however dating the onset of their activity is difficult. The magmatic products of the Galápagos plume, for example, have been subducted and fragmentarily accreted to the Caribbean and South American plates3,4. Based on submarine and terrestrial exposures it is inferred that the plume has been operating for ~90 Myrs5 or perhaps even longer (e.g., ~139 Myrs6). Here we show that the activity of the plume dates back to ~170 Ma. Evidence for this comes from 0 to 168 Ma zircon with isotopic plume signature (Galápagos Plume Array; GPA) recovered from lavas and sediments from ten islands of the archipelago. Given lithospheric plate motion, this result implies that GPA zircon predating the Galápagos lithosphere (i.e., >14 Ma) formed at asthenospheric depths. Thermo-mechanical numerical experiments of plume-lithosphere interaction show that old zircon grains can be stored within local astenospheric stable domains to be later captured by subsequent rising plume magmas. These results open new avenues for research on mantle plume dynamics in similar tectonic settings.

The Pacific plates exposed offshore Central America, Colombia-Ecuador, the Caribbean  Table S4). The zircon grains within this ~0-168 Ma 92 range are distributed continuously without age gaps, defining a "Galápagos plume array" 93 (GPA; Fig. 2a). Zircon younger than 0.2 Ma is rare (steep slope in Fig. 2b), given that 94 most recent lavas are scarcely exposed to erosion. On the contrary, zircon in the range 95 0.2-4 Ma is the most abundant (shallow slope in Fig. 2b), indicating that lavas of that age 96 have the highest exposure to erosion in the different islands. GPA zircon that has pre-97 Galápagos ages in the range ~4-168 Ma is scarce (steep slope in Fig. 2b), but spreads 98 evenly and is isotopically indistinguishable from the younger zircon, suggesting the same 99 plume-related mantle origin.

100
The GPA trend is interrupted at 168 Ma (Fig. 2a) by the appearance of low εHf(t) and high

109
The non-GPA zircon grains indicate old (>213 Ma) to recent (~20 Ma, for the outlier)  The most ground-breaking finding of our extensive zircon study is the group of GPA 118 zircons that pre-date the Galapagos lithosphere and with clear Hf and O isotopic mantle 119 signatures. Given that the age of the Galápagos Islands lavas exposed to erosion is <4 120 Ma (Fig. 2b), that the Galápagos lithosphere is as young as 10-14 Ma 18 and plate motion 121 has removed any older lithosphere from above the plume head, any juvenile GPA zircon 122 older than 14 Ma must have formed in the asthenosphere and have been latter picked up 123 by rising hot-spot magmas at asthenospheric depths (i.e., > ~50 km 28,33 .

124
Two pre-Galápagos GPA zircon grains with ages of ~18 and 22 Ma are slightly younger 125 than the time when the Farallon plate was split by the GSC (just above the plume head) 126 and the Cocos and Carnegie ridges began to form (23 Ma 34 ). This suggests that ridge-127 forming magmas did not fully escape from the plume head and crystallized zircon at 128 asthenospheric depths. The same reasoning can be extended to the other GPA zircon 129 grains older than 23 Ma. During this time, a number of magmatic events took place, 130 including the eruption of the Ecuadorian-Colombian-Caribbean LIP (ECCLIP) with a 131 major phase of LIP construction at ~90 Ma 6 . Notably, we sampled two zircon grains (93 132 and 94 Ma) formed close to this major event (Fig. 2b). The GPA zircon also includes ages 133 younger and older than the major ECCLIP event, clustering at early Tertiary (53, 55 and  Floreana, San Cristobal, Santa Cruz and Santa Fé islands) and inland lava tube deposits 152 (Santa Cruz) clearly point to a local provenance from erosion of exposed volcanic rocks.

153
Furthermore, a provenance study carried out on beaches from eleven islands of the 154 archipelago shows that mineral grains and rock fragments derive from locally exposed  to the surface along with freshly erupting lavas. The zircon age is accordingly the age 193 between formation and eruption (in Ma). Zircon that is transported to depths >300 km has its U-Pb age reset upon reidite formation e.g.,42 and is no longer considered in the 195 interpretation.

196
The model results show that zircon can stay in the shallow upper mantle for extended 197 periods of time (Fig. 3). Counterintuitively, not all zircon is dragged along with the 198 moving lithosphere, but much instead initially moves in the opposite direction. This is 199 because the plume's rising velocity is larger than the plate motion, which induces small 200 scale convective cells (Fig. 3b-c). Some of this zircon returns to the plume area, whilst  Our results thus suggest that, once formed in a plume head, zircon crystals can remain 220 within the asthenospheric mantle for extended periods of time. Following these results, 221 the recorded asthenospheric zircon ages hence allow dating the Galápagos plume back to 222 Jurassic times, a much older age than previously reported. Similar zircon observations 223 and models of asthenospheric flow below ocean islands could apply to other plume-224 related hot spots. Therefore, a systematic analysis of zircon from ocean islands will allow 225 monitoring temporal ranges of plume activity and dynamics over much longer periods 226 than those implied by the ages of the erupted lavas, hotspot tracks, plateaus and, 227 eventually, plume-related terranes accreted to active continental margins.       Figures S4.1, S4.2).

413
Density depends on temperature and volumetric melt fraction: Where ρeff is the effective density, ρsolid is the solid density, ρmelt is the melt density, α is 415 the thermal expansion, and T is the actual temperature and Tref is the reference state 416 temperature. We assume that melt feedbacks on density and viscosity reach their 417 maximum effects at φ=0.08 (see eq. 5) as we do not explicitly account for melt extraction  Table S6. 422 We employ a purely viscous constitutive model in our numerical simulation, using both Bdiff and Bdisl are the pre-exponential factor of diffusion and dislocation creep respectively.

433
α is the pre-exponential factor to simulate the melt viscous weakening, n is the stress 434 exponent, εII is the second invariant of the strain rate tensor, Eact and Vact are the activation 435 energy and volume respectively (see Supplementary Table S6).  initial plume conduit to trigger the upwelling as soon as the simulation starts, which has 454 the same phases as the plume and an excess temperature of 250 °C. Its width is equal to 455 the inflow diameter (i.e. 300 km) and covers almost all the lower mantle with its height.

456
Since, the zircon that have been collected from Galápagos are older than the Nazca plate, 457 we assume that the pacific plate retains its integrity for the whole duration of the  Array that extend from 4 to 168 Ma well beyond the oldest age of the exposed lavas respectively, correspond to zircon grains separated from rock samples.  Isotopic composition (εHf(t) and δ18O(zircon)) and age of Galápagos zircons | A) U-Pb age vs εHf(t) (blue) and δ18O(zircon) (red) of analysed zircons. Both high εHf(t) and low δ18O(zircon) de ne the Galápagos Plume Array (GPA, blue and red rectangles, respectively), which extends from 0 Ma to ca. 170 Ma (note signi cant scatter at >170 Ma). B) Age of analysed zircons sorted by age of spot. The distribution shows four sectors separated by slope breaks, including 1: age range of zircon of less abundant (most recent) volcanic rocks exposed to erosion (<0.2 Ma); 2: age range of zircon of most abundant volcanic rocks exposed to erosion (0.2-4 Ma); 3 and 4: age ranges of pre-Galápagos Islands zircon, comprising 3: zircons belonging to the Galápagos Plume Array that extend from 4 to 168 Ma well beyond the oldest age of the exposed lavas and the age of basement oceanic lithosphere (10-14 Ma), and 4: exotic zircons older than 168 Ma with scattered εHf(t) and δ18O, including continental crust signature.
For reference, 23 Ma (estimated split of Farallon plate) and 90 Ma (major ECCLIP event) are indicated in B. In A and B, lighter coloured circles for εHf(t) and age data, respectively, correspond to zircon grains separated from rock samples.

Figure 3
Numerical models of plume-lithosphere interaction | Snapshots a-c show the dynamics of a plume (with 1600 °C) interacting with a 30 Myrs old oceanic lithosphere that moves with 5 cm/yr to the right. Partially molten regions (indicated by purple shaded areas produce chemically distinct mantle domains (green/violet). We additionally highlight the chemical heterogeneous mantle domain that undergoes the zircon-reidite phase transition. The pathway of zircons in the mantle is tracked by passive tracer zircons (circles) which are coloured by age until a zircon arrives at a partially molten region for the second time when they are assumed are removed from the numerical domain (indicated by stars). Both the age (d) and amount (e) of erupted passive tracer zircons are tracked throughout the simulation and show that old zircons can be preserved in the shallow mantle for extended periods of time.