A record of plume-induced plate rotation triggering subduction initiation

The formation of a global network of plate boundaries surrounding a mosaic of lithospheric fragments was a key step in the emergence of Earth’s plate tectonics. So far, propositions for plate boundary formation are regional in nature; how plate boundaries are created over thousands of kilometres in geologically short periods remains elusive. Here we show from geological observations that a >12,000-km-long plate boundary formed between the Indian and African plates around 105 Myr ago. This boundary comprised subduction segments from the eastern Mediterranean region to a newly established India–Africa rotation pole in the west Indian Ocean, where it transitioned into a ridge between India and Madagascar. We identify coeval mantle plume rise below Madagascar–India as the only viable trigger of this plate rotation. For this, we provide a proof of concept by torque balance modelling, which reveals that the Indian and African cratonic keels were important in determining plate rotation and subduction initiation in response to the spreading plume head. Our results show that plumes may provide a non-plate-tectonic mechanism for large-plate rotation, initiating divergent and convergent plate boundaries far away from the plume head. We suggest that this mechanism may be an underlying cause of the emergence of modern plate tectonics. A mantle plume induced plate rotation that initiated subduction and rifting along a >12,000 km plate boundary about 105 Myr ago, according to an analysis of geological data and numerical simulations.

The formation of a global network of plate boundaries surrounding a mosaic of 23 lithospheric fragments was a key step in the emergence of Earth's plate tectonics. So far, 24 propositions for plate boundary formation are regional in nature; how plate boundaries 25 are created over thousands of kilometers in geologically short periods remains elusive. 26 Here we show from geological observations that a >12,000 km-long plate boundary formed far away from the plume head. We suggest that this mechanism may be an underlying 37 cause of the emergence of modern plate tectonics. 38 The early establishment of plate tectonics on Earth was likely a gradual process that 39 evolved as the cooling planet's lithosphere broke into a mosaic of major fragments, separated by 40 a network of plate boundaries: spreading ridges, transform faults, and subduction zones 1 . The 41 formation of spreading ridges and connecting transform faults is regarded as a passive process, 42 occasionally associated with rising mantle plumes 2 . The formation of subduction zones is less 43 well understood. Explanations for subduction initiation often infer spontaneous gravitational 44 collapse of aging oceanic lithosphere 2 , or relocations of subduction zones due to intraplate stress 45 changes in response to arrival of continents, oceanic plateaus, or volcanic arcs in trenches 3 .

46
Mantle plumes have also been suggested as drivers for regional subduction initiation, primarily 47 based on numerical modeling 4-6 . But while such processes may explain how plate tectonics 48 evolves on a regional scale, they do not provide insight into the geodynamic cause(s) for the 49 geologically sudden (<10 My) creation of often long (>1000 km) plate boundaries including new 50 subduction zones 7 . Demonstrating the causes of plate boundary formation involving subduction 51 initiation using the geological record is challenging and requires (i) establishing whether 52 subduction initiation was spontaneous or induced; (ii) if induced, constraining the timing and 53 direction of incipient plate convergence; (iii) reconstructing the entire plate boundary from triple 54 junction to triple junction, as well as the boundaries of neighboring plates, to identify collisions, 55 subduction terminations, or mantle plume arrival that may have caused stress changes driving 56 subduction initiation. In this paper, we provide such an analysis for an intra-oceanic subduction 57 zone that formed within the Neotethys Ocean around 105 Ma ago, to evaluate the driver of 58 subduction initiation and plate boundary formation. predates upper plate extension that is inferred from spreading records in so-called supra-64 subduction zone (SSZ) ophiolites [8][9][10]11 . Such SSZ ophiolites have a chemical stratigraphy widely 65 interpreted as having formed at spreading ridges above a nascent subduction zone. Several SSZ 66 ophiolite belts exist in the Alpine-Himalayan mountain belt, which formed during the closure of 67 the Neotethys Ocean 12,13 (Fig. 1A) Geological reconstruction of incipient plate boundary 90 The SSZ ophiolites that formed at the juvenile Cretaceous intra-Neotethyan subduction 91 zone are now found as klippen on intensely deformed accretionary orogenic belts (Fig. 1A) that 92 formed when the continents of Greater Adria, Arabia, and India arrived in subduction zones. We 93 reconstructed these orogenic belts ( Fig. 1) and restored these continents, and the Cretaceous 94 ophiolites that were thrust upon these, into their configuration at 105 Ma ( Fig. 1C) (see 95 Methods).

96
The westernmost geological record of the Cretaceous intra-Neotethyan subduction zone 97 is found in eastern Greece and western Turkey, where it ended in a trench-trench-trench triple 98 junction with subduction zones along the southern Eurasian margin 18 . From there, east-dipping 99 (in the west) and west-dipping (in the east) subduction segments followed the saw-toothed shape 100 of the Greater Adriatic and Arabian continental margins (Fig. 1C) and initiated close to it: rocks 101 of these continental margins already underthrusted the ophiolites within 5-15 My after SSZ onto the Indian margin 13,16 (Fig. 1B, C). The Cretaceous intra-Neotethyan plate boundary may 114 have been convergent to the Amirante Ridge in the west Indian Ocean 13 , from where it became 115 extensional instead and developed a rift, and later a spreading ridge, in the Mascarene Basin that 116 accommodated separation of India from Madagascar 13,27,28 (Fig. 1B). The plate boundary ended 117 in a ridge-ridge-ridge triple junction in the south Indian Ocean 13,28 (Fig. 1B).

118
The newly formed Cretaceous plate boundary essentially temporarily merged a large part 119 of Neotethyan oceanic lithosphere between Arabia and Eurasia to the Indian plate. This plate was 120 >12,000 km long from triple junction to triple junction, and reached from 45°S to 45°N, with  the same time as the Indian Ocean-western Neotethys plate boundary formed ( Fig 1C). However, 153 eastward slab pull below Sundaland cannot drive E-W convergence in the Neotethys to the west, 154 and Andaman SSZ extension may well be an expression rather than the trigger of Indian plate 155 rotation. We find no viable plate tectonics-related driver of the ~105 Ma plate boundary 156 formation that we reconstructed here.

157
A key role, however, is possible for the only remaining geodynamic, non-plate-tectonic,       Torque balance modeling -Forces considered here include (i) the push due to plume-521 induced flow in the asthenosphere and (ii) the drag due to shear flow between the moving plate 522 and a deeper mantle at rest (Fig. S1). In the first case, we disregard any lateral variations. Plume-523 induced flow is treated as Poiseuille flow, i.e. with parabolic flow profile, in an asthenospheric 524 channel of thickness hc, radially away from the plume stem. Since at greater distance plume-525 induced flow will eventually not remain confined to the asthenosphere, we only consider it to a 526 distance 2400 km, in accord with numerical results 41 , and consistent with the finding that there is Viscosity is defined such that the force per area is equal to viscosity times the radial gradient of Here hs is an effective thickness of the layer over which shearing occurs, which is calculated 548 below for a stratified viscosity structure, i.e. laterally homogeneous coupling of plate and mantle 549 and which we will set equal to hc for simplicity. Specifically, with Tx being the time-integrated 550 torque acting on a plate rotating an angle w0 around the x-axis 551 552 and Ty and Tz defined in analogy, the torque balance equation can be written 553 554 w0 cancels out when Tx, Ty and Tz are inserted. Integrals used to compute these torques only 555 depend on plate geometry, h0 cancels out in the torque balance, and we can solve for the rotation 556 angle vector w simply by a 3 x 3 matrix inversion. In the more general case, where we do not set 557 hs and hc equal, w is scaled by a factor hs/hc. 558 If a plate moves over a mantle where viscosity varies with depth, then the force per area 559 F/A should be the same at all depths, and the radial gradient of horizontal velocity dv/dz = F/A · 560 1/h (z). If we assume that the deep mantle is at rest (i.e. it moves slowly compared to plate 561 motions), we further find that plate motion is

562
(1) 563 The integration is done from the base of the lithosphere z0 to the depth where the approximation 564 of the "mantle at rest" is probably the most closely matched, i.e. we choose the viscosity maximum. The last equality is according to the definition of the effective layer thickness, 566 whereby h0 is the viscosity just below the lithosphere. Solving this equation for hs for the 567 viscosity structure in Extended DataFig. 2 and a 100 km thick lithosphere gives hs=203.37 km.

568
The plume location at 27.1°E, 40.4° S, is obtained by rotating the center of the 569 corresponding LIP at 46° E, 26° S and an age 87 Ma (adopted from Doubrovine et al. 73 ) in the 570 slab-fitted mantle reference frame 45 , in which also the plate geometries at 105 Ma are 571 reconstructed.

572
Results for this case ( Fig. 2A) show that a plume pushing one part of a plate may induce 573 a rotation of that plate, such that other parts of that plate may move in the opposite direction. A 574 simple analog is a sheet of paper pushed, near its bottom left corner, to the right: Then, near the 575 top left corner, the sheet will move to the left. With two sheets (plates) on either side, local 576 divergence near the bottom (near the plume) may turn into convergence near the top (at the part 577 of the plate boundary furthest away from the plume). The length of that part of the plate 578 boundary, where convergence is induced may increase, if one plate is nearly "pinned" at a hinge 579 point slightly NE of the plume, perhaps due to much stronger coupling between plate and mantle.

580
At the times considered here ~105 My ago, the Indian continent, where coupling was presumably 581 stronger, was in the southern part of the Indian plate, whereas in its north, there was a large 582 oceanic part, with presumably weaker coupling. Hence the geometry was indeed such that 583 convergence could be induced along a longer part of the plate boundary.

584
In the second case, we therefore consider lateral variations in the coupling between plate 585 and mantle, corresponding to variations in lithosphere thickness and/or asthenosphere viscosity, 586 by multiplying the drag force (from the first case) at each location with a resistance factor. This 587 factor is a function of lithosphere thickness reconstructed at 105 Ma. On continents, thickness 588 derived from tomography 74 with slabs removed 75 is simply backward-rotated. In the oceans, we 589 use thickness [km] = 10 · (age [Ma] -105) 0.5 with ages from present-day Earthbyte age grid 590 version 3.6, i.e. accounting for the younger age and reduced thickness at 105 Ma, besides

Extended Data
Extended Data Fig. 1: Sketch illustrating the geometry of a plume head (pink; not drawn to scale) hitting the boundary of a plate (green). xplu (red arrows) are the (maximum) displacement vectors in the asthenosphere caused by emplacement of the plume. Motion vectors of the plate xpla (black arrows) correspond to the plate rotation ! that is caused.