Hemispheric geochemical dichotomy of the mantle is a 1 legacy of austral supercontinent assembly and deep 2 subduction of continental crust 3 4

Oceanic hotspots with extreme enriched mantle radiogenic isotopic signatures—including high 87Sr/86Sr and low 143Nd/144Nd indicative of ancient subduction of continental crust—are restricted to the southern hemispheric mantle. However, the mechanisms responsible for concentrating subducted continental crust in the austral mantle are unknown. We show subduction of sediments and subduction eroded material, and lower continental crust delamination, cannot generate this spatially coherent austral domain. However, late Neoproterozoic to Paleozoic continental collisions—associated with the assembly of Gondwana and Pangea—were positioned predominantly in the southern hemisphere during the late Neoproterozoic appearance of widespread continental ultra-high-pressure (UHP, >2.7 gigapascals) metamorphic terranes, which marked the onset of deep subduction of upper continental crust. We propose that deep subduction of upper continental crust at ancient rifted-passive margins during austral supercontinent assembly, from 650-300 Ma, resulted in enhanced upper continental crust delivery into the southern hemisphere mantle. In contrast, EM domains are absent in boreal hotspots, for two reasons. First, continental crust subducted after 300—when the continents drifted into the northern hemisphere—has had insufficient time to return to the surface in plumes feeding northern hemisphere hotspots. Second, before the appearance of continental UHP rocks at 650 Ma, upper continental crust was not subducted to great depths, thus precluding its subduction into the northern hemisphere mantle during the Precambrian when continents may have been located in the northern hemisphere. Our model implies a recent formation of the austral EM domain, explains the geochemical dichotomy between austral and boreal hotspots, and may explain why austral hotspots outnumber boreal hotspots.

1 Department of Earth Science, University of California, Santa Barbara, 93106, USA 6 (*corresponding authors: jackson@geol.ucsb.edu, francism@ucsb.edu) 7 8 9 10 11 12 13 This paper is a non-peer reviewed preprint submitted to EarthArXiv. 14 15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43 Introduction temperature/high pressure terranes (14) (including UHP terranes) that introduced continental 141 material into the mantle. A full description of each of these parameters is provided in Methods. 142 Below we propose that the primary path for the delivery of continental crust to the mantle is 143 through the deep subduction of formerly rifted passive margins during arc-continent and 144 continent-continent collisions. Sometime during collisional metamorphism, crust on the lower 145 plate of the collision will detach or break-off (Fig. 3). This crust includes upper and lower 146 continental crust, oceanic crust, transitional crust located between oceanic and continental crust 147 (including significant mafic material), and overlying metamorphosed sediments. As discussed 148 below, some of this crust will enter the mantle and some will be relaminated to the upper plate 149 (17,18). Below we show that, from the late Neoproterozoic to the Paleozoic, continental 150 assembly occurred in the southern hemisphere during onset of deep continental crust subduction, 151 which was made possible by onset of a global late Neoproterozoic transition from shallow to 152 deep slab-breakoff. The coincident timing of continental geography and a global tectonic 153 transition set the stage for focused deep subduction of a formerly rifted passive margins-154 subducted at continent-continent and arc-continent collision zones-into the southern 155 hemisphere mantle from the late Neoproterozoic through the Paleozoic. 156 hotspot geochemistry (low hotspot 3 He/ 4 He) cannot be used to reliably exclude a plume origin 187 for such hotspots (12). Instead, we posit that seismic methods are better suited for determining 188 whether a plume exists beneath a hotspot: at least 28 hotspots have been found to be associated 189 with plumes detected using seismic methods (20,21), including key Dupal hotspots excluded by 190 Doucet et al. (19). Furthermore, plumes beneath some hotspots may be undetected because of the 191 resolution of existing global tomographic models and limited data coverage (12). For example , 192 the Yellowstone plume was not detected in global seismic models, but was detected in a regional 193 study that employed the high-density US array (22), suggesting that we cannot exclude the 194 presence of a plume beneath a hotspot simply because it remains undetected in existing 195 (relatively low resolution) global seismic models. Therefore, we refrain from excluding oceanic 196 hotspots from our treatment, and this approach leads us to observe a clear hemispheric 197 dichotomy in the distribution of EM domains . 198 Further highlighting the northern versus southern hemisphere dichotomy among global 199 hotspots, there are twice as many southern hemisphere hotspots as northern hemisphere hotspots 200 ( Fig. 4): the analysis shown in Fig. 4 includes both oceanic and continental hotspots (i.e., all 47 201 oceanic and 11 continental hotspots shown in Dataset S1). If only oceanic hotspots are counted , 202 there are still more than twice as many southern hemisphere hotspots as northern hemisphere 203 hotspots. However, the mechanism responsible for this hemispheric dichotomy in hotspot 204 distribution is unknown (12). 205 206

Flux and composition of continental crust into the mantle and Dupal formation. The EM 207
domain, sampled by plume-fed OIB, is widely suggested to form by a contribution from 208 subducted continental crust. However, the exact type of continental material that contributes to 209 the EM mantle-upper continental crust, lower continental crust, or sediment derived from 210 continental crust-is not well constrained (2,(23)(24)(25)(26). Below we discuss the different mechanisms 211 for introducing continental crust into the mantle and how the geographically restricted 212 distribution of the Dupal domain constrains the origin, and type, of continental crust material that 213 makes up the EM domain. 214 As outlined in Stern and Scholl (27), continental crust is subducted in one of four 215 principal ways: 1) sediment subduction results in upper continental crust loss, 2) subduction 216 erosion removes upper and lower continental crust, 3) lower crustal foundering removes lower 217 continental crust only, and 4) wholesale continental crust subduction removes upper and lower 218 continental crust. However, the absolute fluxes of continental crust material into the mantle that 219 result from each of these mechanisms are highly uncertain in the present day, as widely 220 acknowledged in work dedicated to the subjectHacker11_Hacker15_Stern&Scholl10_Clift09 221 (17,18,27,28). For example, as discussed in Hacker et al. (18), Scholl & von Huene (29) 222 estimated that 95% of subducted sediment is returned to the mantle, but Hacker et al. (17) argue 223 that the low density of sediment will result in most (up to 94%) of the subducted sediment 224 simply being relaminated to the underside of the continental crust. Thus, the relamination 225 mechanism hinders sediment delivery into the convecting mantle. Similarly, subduction erosion 226 also has been argued to be efficient at transporting continental crust into the mantle (27,28), but 227 Hacker et al. (17) argue that, due to having relatively low density, most (up to 82%) of the 228 products of continental crust subduction erosion are simply relaminated onto the underside of the 229 continental crust instead of being returned to the mantle. In summary, the low density of 230 sediment and the products of subduction erosion may result in relamination, which greatly 231 reduces loss of this continental crust-derived material to the mantle (17,18) However, the long-term survival of Hadean 142 Nd, 182 W, and 129 Xe anomalies identified in 250 modern OIB (1,30,31) suggests that the mantle is not an efficient blender. Therefore, it seems 251 difficult to argue that, once subducted, continental crust and its geochemical signatures are easily 252 attenuated in the mantle over time. 253 As with subduction of sediments and the products of subduction erosion, the geographic 254 distribution of the Dupal domain can also be used to exclude an origin by lower continental crust 255 delamination. Lower continental crust is widely assumed to be mafic in composition (e.g., (32)) 256 (but see Hacker et al. (17,18) for an argument that most-up to ~80%-of the lower continental 257 crust is silicic). Therefore, in contrast to relamination of low-density sediments and products of 258 subduction erosion, delamination of dense mafic lower continental crust results in continental 259 crust loss to the mantle (e.g., (33,34)). Lower continental crust delamination may have been 260 operating since the Archean (e.g., (35)) and, like subduction zones, the continents have been 261 located over a wide range of latitudes since the Archean, including the northern hemisphere. 262 Therefore, if lower continental crust delamination were responsible for generating EM domains, 263 EM mantle reservoirs would be expected at all latitudes, including outside of the southern 264 hemisphere Dupal domain, but this is not observed. A conclusion of this observation is that 265 delaminated lower continental crust does not generate EM domains. There are several possible 266 reasons why this would be the case. First, it is possible that delaminated mafic lower continental 267 crust is too dense to return to the surface in upwelling plumes that source hotspots. However, if 268 lower continental crust does return to the surface, it is possible that lower continental crust does 269 not have geochemical characteristics necessary to incubate EM domains during long-term 270 storage in the mantle. Alternatively, if lower continental crust is more silicic in composition, as 271 suggested by Hacker et al. (17,18), then lower continental crust may be insufficiently dense to 272 enter the mantle in the first place, so is incapable of generating deep mantle EM domains sourced 273 by plumes. 274 Nonetheless, positive Eu anomalies (i.e., Eu/Eu*>1; see Fig. S1 for definition) in OIB 275 have been used to argue that lower continental crust, which can have Eu/Eu*>1, contributes to 276 some OIB with EM1 (enriched mantle I) signatures (25,26). However, we find that upper and 277 lower continental crustal rocks both exhibit examples where Eu/Eu*>1 (Fig. S1), and thus 278 Eu/Eu*>1 is not diagnostic of lower continental crust in the mantle sources of OIB. Furthermore, 279 high Eu/Eu* values in OIB may not reflect their mantle sources due to diffusive interaction with 280 plagioclase-rich lithologies in the lower oceanic crust (36)  spatially restricted to the southern hemisphere (6) (Fig. 1). 290 291

Continental crust subduction at rifted passive margins and attendant UHP metamorphism.
In 292 contrast to subducting sediments and products of subduction erosion, the continental and 293 transitional crust along rifted-passive margins are attached to dense down-going mafic slabs ( Fig.  294 3), which confer greater effective density to the subducting package (i.e., the denser oceanic slab 295 is attached to less dense continental crust, and together they form a package with higher density 296 than upper continental crust alone (16) (Fig. 3). In this way, continental crust-including silicic 297 upper continental crust and transitional crust, where the latter is sandwiched between continental 298 and oceanic crust and is composed of silicic and mafic rocks-can descend into the mantle as 299 long as it is attached to the dense oceanic slab. Once the subducting package reaches the point-300 of-no-return-a mantle depth below which the silicic portion of continental crust undergoes 301 phase transitions to become denser than ambient mantle (17)-the silicic continental crust will 302 continue to sink, even if it detaches from the down-going slab. However, as discussed below, if 303 the slab detaches from the continental crust before reaching the depth of no return, the silicic 304 continental crust will simply rise buoyantly. The feasibility of subduction of silicic upper 305 continental crust compositions to depths deeper than the point of no return has been 306 demonstrated with geodynamic modeling (15,35), which shows that silicic continental crust can 307 achieve densities greater than ambient mantle at the point of no return. In contrast to continental 308 crust attached to down-going slabs at ancient rifted-passive margin settings, sediments and the 309 products of subduction erosion are not attached to a down-going slab, so they are buoyant and 310 simply relaminate and, therefore, are not conveyed to great depths. 311 Of course, both upper and lower continental crust are attached to the down-going slab 312 and will be subducted together during collision. However, as described above, lower continental 313 crust-which has been delaminated at all latitudes since the Archean (35)-cannot explain 314 formation of the of the geographically restricted Dupal domain. Given that delaminated and 315 subducted lower continental crust should have broadly the same composition, we see no reason 316 why lower continental crust delivered to the mantle during continental subduction would 317 contribute to the Dupal domain when delaminated lower continental crust does not. In short, we 318 do not know the fate of lower continental crust after it enters the mantle, but it is not a good 319 candidate for generating EM domains for reasons described above. 320 Instead, it is the silicic upper continental portion of the subducted ancient rifted-passive 321 margin that is likely to contribute to Dupal formation. This is supported by geochemical 322 evidence consistent with upper continental crust signatures in the EM mantle domain (24,25). 323 We acknowledge that the origin of the geochemical signatures in EM1 (24) and EM2 (25) 324 previously were attributed to the addition of subducted marine sediments to these mantle sources, 325 but much of the sediment budget in the ocean basins (both terrigenous and marine sediment 326 (38,39) (6,10). During this period, the 357 continents moved deep into the southern hemisphere, and were positioned, on average, south of 358 15° south latitude ( Fig. 5; Dataset S2). This time interval-approximately 650 to 300 Ma-359 defines a peak in the subduction of ancient rifted-passive margins in Earth history, which was 360 focused in the southern hemisphere ( Fig. 5 panel b). 361 An important question is whether the paleogeography of the continents relative to the 362 paleomagnetic (or spin-axis) reference frame, like that used in Merdith et al. (43), can be used to 363 constrain the latitude of subducted contributions to the mantle. Prior to the Mesozoic, absolute 364 paleogeography is difficult to determine because paleomagnetic data constrains only latitude, not 365 longitude, and because the paleomagnetic reference frame can move relative to a mantle 366 reference frame (i.e., true polar wander). The Neoproterozoic reconstruction of Merdith et al. 367 (43) uses a paleomagnetic reference frame, and relies on geological and kinematic constraints, 368 but is merged into Paleozoic models based on a hybrid mantle reference frame (44,45). A large 369 Paleozoic longitudinal shift in the reference frame is necessary to reconcile these models (43). 370 Although it would be ideal to use a mantle reference frame, going back to 1000 Ma, to present 371 the Merdith et al. (43) reconstruction, this does not exist. Moreover, there is disagreement 372 whether to use a mantle reference frame with true polar wander around hypothesized fixed 373 LLSVPs (42,46), or if the LLSVPs are mobile on 100 Myr timescales (47). To assess the 374 sensitivity of reference frames on paleolatitude, we compared paleogeographic reconstructions at 375 100 Myr snapshots in the mantle versus paleomagnetic reference frames in the reconstruction of 376 Torsvik & Cocks (2017) (42), which extends to 600 Ma (Fig. S2). We find that a southern 377 hemisphere dominance of continental latitudes persists in both the paleomagnetic and the 378 modeled mantle reference frame from 600-300 Ma (i.e., the proposed interval of continental 379 crust contribution to the southern hemisphere mantle). Inferring paleolatitudes prior to ~550 Ma 380 is less straightforward because paleomagnetic studies from ca. 582 to 565 Ma rock units have 381 yielded anomalous results (e.g., (48,49)), which have been interpreted to record large-scale true 382 polar wander (e.g., (13,50). However, recent studies suggest that the Ediacaran field was 383 anomalously weak and unstable with high-frequency reversals (e.g, (51,52)). Therefore, a simple 384 interpolation of the paleomagnetic data through this interval (e.g., (43)), with the assumption of 385 no significant true polar wander or difference between paleolatitude in the paleomagnetic and 386 mantle reference frames (Fig. S2), appears to be the most parsimonious approach, supporting a 387 southern hemisphere positioning of the continents from 650 to 550 Ma (Fig. 5). 388 The pulse of ancient rifted-passive margin subduction from 650-300 Ma is illustrated in 389 Fig. 5 (panel b), which shows that subduction of these margins occurred overwhelmingly in the 390 southern hemisphere from 650-300 Ma. Critically, this maximum in ancient rifted-passive 391 margin subduction events from 650-300 Ma occurred on the heels of the breakup of Rodinia, a 392 breakup marked by a peak in the generation of rifted-passive margins (Fig. 5), which exposed 393 Archean and Proterozoic cratonic and transitional crust on margins (53) (Fig. 3). During 394 Gondwana and Pangea construction, many of these ancient rifted-passive margins were on the 395 lower, down-going plate of collisional orogens (see (54)) which provides many examples of 396 Archean to Proterozoic crust along margins on the lower plate during the assembly of 397 Gondwana), setting the stage for their subduction into the mantle during collision. Additionally, 398 UHP metamorphic terranes with continental protoliths-which record continental crust delivery 399 to great depths-can provide long-term records of deep continental subduction; the reconstructed 400 paleolatitude of continental UHP terranes shows that, from at least 600-300 Ma (42) (Fig. S2 While we propose that the period from 650-300 Ma represents a major episode of upper 411 continental crust subduction in Earth's history, we emphasize that continental subduction of 412 ancient rifted-passive margins prior to 650 Ma-when continents were also in the northern 413 hemisphere-did not efficiently deliver continental material to the mantle: while Precambrian 414 collisional orogens likely occurred over a wide range of latitudes that included the northern 415 hemisphere, these events did not deliver upper continental crust past the point of no return, 416 otherwise they would have also generated EM domains in the northern hemisphere of the Earth's 417 mantle, which is not observed. This hypothesis is supported by observations from the rock 418 record, which shows that continental crust did not reach depths required for UHP metamorphism 419 until the ~650 Ma (i.e., the oldest known UHP metamorphic rock with a continental protolith 420 (14)). This is not a new result: for example, van Hunen and Allen (16) write that, prior to late 421 weakening of the subducting slab in the higher geothermal gradient of the Precambrian Earth 438 favored shallow slab-breakoff of the denser oceanic lithosphere (15). During this time interval, 439 the dense slab simply detached from down-going continental crust at shallow depths before 440 continental crust could reach UHP conditions (15), and the continental crust did not reach the 441 point of no return (Figs. 3 and 6). Following detachment, the up-dip felsic subduction complex 442 had higher effective buoyancy (Fig. 3), which permitted felsic continental crust to migrate back 443 up the subduction channel (55) before reaching before reaching UHP conditions. This model is 444 consistent with the absence of continental UHP metamorphism (i.e., absence of continental 445 protoliths with pressures >2.7 GPa; Fig. 6) prior to 650 Ma (15,16). 446 The appearance of widespread continental UHP metamorphism in the late 447 Neoproterozoic is proposed to result from evolution of the style of subduction in response to 448 secular cooling of the Earth's interior: down-going slabs began to break off from the continental 449 crust at greater depths due to a rheological strengthening of the lithosphere as the mantle cooled 450 ( Fig. 3) (14,15,56). Because felsic continental crust is attached to the denser down-going slab to 451 greater depths, the up-dip felsic subduction complex could be transported to greater depths (55), 452 permitting generation of continental UHP rocks for the first time (Fig. 6, panel b). While 453 continental crust observed at UHP metamorphic terranes was exhumed, continental crust 454 subducted past the point of no return was delivered into the mantle (15,27,56). From ~650 Ma 455 (i.e., the age of the oldest known UHP metamorphic rock with a continental protolith (14)) to 456 present, zones of continent collision were nurseries for continental UHP metamorphic terranes 457 that enhanced continental crust delivery to mantle depths (but, as discussed below, continental 458 crust subducted from 300 Ma to present has not had sufficient time to cycle back into the sources 459 of hotspots). 460 While fluxes of continental materials into the mantle are poorly constrained in the present 461 day, the widespread appearance of continental UHP metamorphic terranes by 650 Ma provides 462 evidence that the evolving thermal evolution of the Earth became more conducive to delivery of 463 continental material to great depths (27,56). This rheological transition within the slab-which 464 enabled deeper continental subduction-may not have been abrupt, but could have occurred on a 465 longer timescale (i.e., following Rodinia) and became evident, and widespread, only during 466 abundant continent collisions that occurred in the late Neoproterozoic (14). We propose that deep  continents moved northward over the past 300 Ma (Fig. 5, panel b). Therefore, a key question is 505 why continental crust subduction from 300 Ma to present did not generate a deep EM domain 506 sampled by northern hemisphere hotspots. 507 We argue that subduction of continental crust into the northern hemisphere during the last 508 300 Ma was unlikely to generate extreme EM domains sampled by modern hotspots (12) largely 509 because continental material subducted into the northern hemisphere over the past several 510 hundred million years has not had sufficient time to return to the surface in northern hemisphere 511 mantle plumes. Down-going slabs require ~200 Ma to reach the CMB (44,60), then they reside at 512 the CMB for a period of time, and finally they need 10's to 100 Ma to be transported from the 513 CMB to the near surface in upwelling plumes (46,61). Therefore, northern hemisphere oceanic 514 hotspots do not exhibit geochemical signatures associated with northern hemisphere continental 515 crust subduction that occurred from 300 Ma to present (12). 516 517

Continental crust in the austral regions of the Large Low Shear-wave Velocity Provinces gives 518
rise to more radiogenic heating, thereby generating more austral plumes. The distribution of 519 the 11 EM hotspots in the southern hemispheric mantle is not random: all are geographically 520 restricted to the southern hemispheric regions of the two deep mantle Large Low Shear-wave 521 Velocity Provinces (LLSVPs) (Fig. 1). The clear geographic correspondence between LLSVP 522 structures at the bottom of the mantle and EM hotpots at the surface is strong evidence that the 523 material hosting the EM signatures-deeply subducted upper continental crust-resides in the 524 deep mantle LLSVP structures (8,12). If we are to accept that upper continental crust can be 525 transported past the point of no return during low T/P metamorphism recorded at UHP terranes, a 526 clear implication of the geographic link between the Dupal domain and the LLSVPs is that the 527 subducting continental crust enters the deepest mantle after passing the point of no return. How 528 deeply subducted continental crust preferentially enters the LLSVPs, which cover only ~30% of 529 the core-mantle boundary (62), remains a key question. 530 The composition, temperature, density structure, and origin of the LLSVPs are the 531 subject of intense debate: they may host dense oceanic crust or primordial material, or both, 532 which is consistent with suggestions that the LLSVPs have higher density and distinct chemical 533 makeup (e.g., (62,63,64)). The clear geographic correspondence between EM hotspots and the 534 LLSVPs leads us to argue that the LLSVPs are also host to most of the subducted upper 535 continental crust in the deep mantle. Below we argue that this is a consequence of convection 536 patterns in the mantle. 537 The global geographic anticorrelation between subduction zone locations and LLSVPs, 538 where no slabs intersect with the LLSVPs (44), suggests that the mantle is dominated by a broad 539 convection pattern with subduction zone downwelling around the LLSVPs and upwelling above 540 the LLSVPs: a consequence of this convection pattern (20,65,66) is that down-going slabs will 541 tend to push deep compositional layers-including subducted continental crust (12) We note that the quantity of subducted continental crust incorporated into the LLSVPs is 563 not known, owing to great uncertainty associated with fluxes of continental crust to the mantle. 564 Additionally, the observation of proportionately more austral hotspots could relate to the fact that 565 the center of mass of the LLSVPs-which appear to source most hotspots (Figs. 1 and 4)-are 566 shifted into the southern hemisphere (12). Nonetheless, we calculate that the heat flow carried by 567 plumes (which is at least 2.0±0.3 TW (67)) can be matched by a deep continental reservoir that 568 with a mass of continental crust that is ~28% of the modern continents, well within the range of 569 prior estimates for total subducted upper continental crust materials over geologic time (68). 570 Therefore, if located primarily in the southern hemisphere portions of the LLSVPs, a relatively 571 modest contribution from a deeply subducted upper continental crust reservoir in the southern 572 hemisphere could provide additional heat to explain both the greater number of hotspots and the 573 higher summed hotspot buoyancy flux in the southern hemisphere. 574 575

Methods 576
Geochemical database of oceanic hotspot lavas. The hotspot geochemistry database includes the 577 lowest 143 Nd/ 144 Nd (12) lava-here called the most EM sample-analyzed at each of the 578 geochemically characterized oceanic hotspots, and is shown in Fig. 2. The geochemical database 579 was published previously (12), where an extensive description of data curation is also provided. 580 Nonetheless, a description of data curation in the geochemical database is provided in the SI 581 Appendix. Additionally, a list of the 58 known hotspots (47 oceanic and 11 continental hotspots), 582 which we examine here, is provided in Table S1 (we note that one oceanic hotspot, Vema, has 583 not been geochemically characterized, thus only 46 oceanic hotspots are geochemically 584 characterized and shown in Fig. 2). The geochemical data in the database is compiled in Jackson 585 et al. (12), together with sources of the data, and the database is not republished here. 586 587 Plume locations beneath hotspots. In order to identify the deep mantle location of the enriched 588 mantle domains sampled by each hotspot, we do not use the surface location of the hotspots and 589 project vertically downward. Instead, we embrace the fact that plume conduits tilt as they rise 590 through the mantle (20,21,61). Therefore, when examining the geographic distribution of 591 geochemical domains in the mantle, we use the latitude of the calculated conduit bases at 2850 592 km depth beneath each hotspot, near the core-mantle boundary (CMB) (12). Methods for plume 593 advection use here are published elsewhere (12), but are nonetheless are provided in the SI 594 Appendix. An average of plume conduit locations for each hotspot at the base of the mantle, and 595 hotspot locations at the Earth's surface, are calculated from individual conduit locations from 596 Jackson et al. (12) and are provided in Table S1, as described in the SI Appendix. 597 598 Ancient rifted-passive margins over the past billion years. A previous compilation of passive 599 margin formation, tenure, and death through Earth History (69) accounted for the number of 600 rifted-passive margins and their terminations through time, but not the geographic length, 601 latitude, or the timeframe between the start and end of collisional metamorphism. In order to 602 calculate the lengths, locations (latitudes), and duration of subducted rifted-passive margins over 603 the past 1000 Ma, individual margins were modified from previous compilations (69,70) and 604 traced from georeferenced geological maps in QGIS (as shown in Fig. S3). The margins were 605 captured at a similar spatial resolution (i.e., node spacing). These rifted-passive margin lines 606 were assigned a start of rifting age, a passive margin start-date at the rift-drift transition, a 607 passive margin end-date (i.e., at the beginning of collision, or "death" in Bradley (69) The description of each of the rifted-passive margins in our database is provided in the SI 614 Appendix. As we are primarily concerned with the terminations of these margins, we lump 615 composite rifts associated with failed rifts or rifting of a ribbon continent together as one margin. 616 These processes can be discerned in younger examples, but are more difficult to define in older 617 margins, which may contribute to the apparent secular decrease in rifted-passive margin lifespan 618 (69). Most rifted-passive margins will end with an arc-continent collision. As such, sutures are 619 assigned to each margin and the rifted-passive margin end-date corresponds with the start of 620 exhumation date (70). 621 A collision begins when an ancient rifted-passive margin, with crust of thickness and 622 composition that is transitional between oceanic and continental crust, enters the trench. 623 Eventually this transitional and continental crust will jam the subduction zone, resulting in a 624 collisional orogeny, and the subducting slab including some transitional and continental crust 625 will break-off into the mantle (71). We define a collision end date to encompass the duration of 626 subduction of transitional and continental crust to the mantle. We define the end of collision 627 where constraints are known from reorganization of the plate geometry, geophysical imaging, or 628 metamorphic ages. However, in most orogens that encompass an arc-continent collision followed 629 by a continent-continent collision, like the India-Asia example, crustal thickening can persist for 630 tens of millions of years after the initial collision (72). On the Himalayan margin, arc-continent 631 collision began by 52 Ma along the Indus-Tsangpo and Shyok sutures (73), marked by the 632 appearance of young volcanic material on the Indian margin (74). However, after the initial arc-633 continent collision, the subduction zone stepped north and consumed the Kshiroda plate (75,76)  where other constraints are lacking. For continent-continent collisions, we do not count the 644 length of both margins, but instead count only the lower plate in the collision. Many collisional 645 margins lack evidence for deep continental subduction, such as the Thor suture between Baltica 646 and Avalonia; however, to minimize subjective filters, issues of preservation bias, and to keep 647 the compilation independent from the UHP database (14), we preserve these margins in the 648 database. 649 Some rifted-passive margins, such as the Pyreneean-Biscay margin, transformed into an 650 active arc and retro-arc foreland without evidence for a collisional phase (69). These margins 651 were relatively young and may have initiated subduction along transcurrent margins, and as such 652 there is no evidence for subduction of the margin. Consequently, we have eliminated these types 653 of margins from our compilation. We also do not include proposed rifted-passive margins (69) in 654 the Farewell Terrane, the Hoggar, or the Idermeg terrane of Mongolia, because geological 655 constraints on these terranes are lacking. Therefore, our estimates of subduction of 656 Neoproterozoic rifted-passive margins is a conservative minimum. Additionally, although 657 several Gondwanan sutures are present in Antarctica, too little is known to define Proterozoic 658 rifted-passive margins and these are also excluded from our treatment. 659 The paleolatitude of rifted-passive margin subduction from the beginning to end of 660 collision was determined by assigning a plate ID to each margin and restoring to its position at 661 the time of collision onset in GPlates (43). For each margin, the latitude was extracted from the 662 latitudinal midpoint at 5º resolution (using data from a paleomagnetic reference and spin-axis 663 frame (43) with modification from recent literature (77) 188-197 (2010).

Supplementary Information Text Geochemical database of oceanic hotspot lavas.
The geochemical database providing the most extreme (lowest 143 Nd/ 144 Nd) lava from all known geochemically-characterized oceanic hotspots is published elsewhere (1). To explore the distribution of EM domains, we examine only the lowest 143 Nd/ 144 Nd lava from each hotspot. Alternative approaches, such as identifying the mean or median 143 Nd/ 144 Nd of all lavas from each hotspot, do not identify the global distribution of the most enriched geochemical domain sampled by each hotspot.
As discussed elsewhere (2), the minimum hotspot 143 Nd/ 144 Nd is the preferred indicator of geochemical enrichment in oceanic lavas over 87 Sr/ 86 Sr and 176 Hf/ 177 Hf because 1) 143 Nd/ 144 Nd is less susceptible to seawater contamination than 87 Sr/ 86 Sr, and 2) far more 143 Nd/ 144 Nd data are available in OIB than 176 Hf/ 177 Hf data. Derived Pb-isotope parameters (i.e., 208 Pb*/ 206 Pb*, Δ 207 Pb/ 204 Pb, Δ 208 Pb/ 204 Pb) (3, 4) also correlate with geochemical enrichment, but Pb isotope databases for OIB suffer from lower precision datasets that were generated prior to the advent of modern techniques that monitor in-run Pb isotope fractionation (e.g., MC-ICP-MS analyses using Tl addition or double-and triple-spike Pb isotope analysis by TIMS). Available Pb isotopic data for much of the oceanic hotspot dataset was obtained using older TIMS methods that did not control in-run isotope fractionation, leaving many hotspots without available high-precision modern Pb isotope measurements. Therefore, we do not further explore Pb isotopic data here. We note that Hart (3) (Fig. 2).
The geochemical database used here and description of methodology for database construction are provided in (1). Hotspot lavas erupted in continental settings are excluded from the analysis because continental crust assimilation-a mechanism that can impart low 143 Nd/ 144 Nd and high 87 Sr/ 86 Sr on upwelling mantle-derived melts-can mask mantle signals with shallowderived continental fingerprints. Therefore, the geographic extent of the Dupal domain is defined using oceanic hotspots only, and the histogram in Fig. 4 shows the global latitude distribution of all hotspots, oceanic and continental. In the published geochemical database used here, Sr and Nd isotopes were obtained on the same sample (i.e., the sample with the lowest 143 Nd/ 144 Nd from each hotspot). We do not apply an "age threshold" for the analysis: we assume that a volcano located along a particular hotspot track erupted over the hotspot, no matter the age of the volcano, and this approach allows us to assign the latitude and longitude of the hotspot to samples collected anywhere along a hotspot track.

Plume locations beneath hotspots.
Because it is difficult to resolve plume conduits under some hotspots due to the low resolution available in global seismic models, plume conduits calculated in plume advection models (5-7) allow us to explore how the different plumes tilt as they up-well, and infer the location of the plume conduit at the core-mantle boundary.
Methods for the calculation of advected conduit bases presented here are presented in (1), together with hotspot locations at the surface. In short, to define plume conduit locations beneath each hotspot, we use an average location of the calculated advected conduits bases calculated at 2850 km depth from four seismic models: SEMUCB-WM1 (8), S40RTS (9), SMEAN2 (10), and TX2016 (11). The average of the plume conduit locations at 2850 km from the four models is shown in Dataset S1. The variability in the latitude of the conduit base location across the four seismic models is reflected in the 2 SD error bars on latitude in Fig. 2 (see Dataset S1). Although using the latitude of the surface location of the hotspot would not significantly change the results of this study, our treatment is more accurate because plumes are advected laterally in the convecting mantle as they rise (6)(7)(8). Advected conduits show that the deep mantle plume source for two of the 47 oceanic hotspots-Hawaii and Caroline hotspots-are in a different hemisphere than the surface location of the hotspot in some seismic models. First, the surface location of the Hawaiian hotspots is located at 19° N at the Earth's surface, but the calculated plume conduit base is located in the southern hemisphere in plume advection models run in two global seismic models (SEMUCB-WM1 and TX2016) and in the northern hemisphere (but near the equator) in two other seismic models (SMEAN2 and S40RTS) (see plume conduit results in Jackson et al. (1)). Nonetheless, the latitude of the base of the plume conduit beneath Hawaii overlaps with the southern hemisphere within 2 SD uncertainty (see Fig. 2 and Dataset S1). The surface location of the Caroline hotspot, at 5° N latitude, has a calculated conduit base that is located in the southern hemisphere in all plume advection models (1). All of the other hotspots explored in the study have plume bases that are calculated to reside in the same hemisphere as the modern surface expression of the hotspot.

Passive margins of the past billion years.
The following text provides a description for ancient rifted-passive margins used in our database. The ancient rifted-passive margins are labeled with a number (ID) that is used to identify the margin on the map (Fig. S3) and in Dataset S3.

Innuitian
Thick Late Mesoproterozoic to early Tonian platformal strata define the northern margin of Laurentia. Based on new dates and stratigraphy (12,13), the Mesopoterozoic Borden-Bylot basins can now be linked to a series of intercontinental basins in northern Laurentia and Siberia, and have consequently been removed from the compilation.
The northern margin of Laurentia was intruded by the 719 Ma Franklin large igneous province (LIP) (14), which was associated with the separation of Siberia from Laurentia. However, the Innuitian margin does not preserve a Cryogenian to Ediacaran rifted passive margin. Late Ediacaran mixed carbonate and siliciclastic rocks of the Kennedy Channel Formation are present on Ellesmere Island (15), which may record reactivation of the margin. Broad deposition across the Arctic margin does not occur until after the early Cambrian (16), which we attribute to thermal subsidence, and place the rift-drift transition at 525 Ma (17).
The passive margin became a foreland basin with collision of the McClintock arc to the east (16). The earliest stratigraphic record of collision is a latest Ordovician to earliest Silurian (ca. 445 Ma) influx of orogen-derived turbidites in northernmost Greenland and Ellesmere Island (18). Foreland deposits persist through the Silurian. The suture occurred only in the NE segment and was translated modified by later sinistral motion along the margin. We extend the subduction of crustal material to the mantle for 12 Myr to 433 Ma along this arc-continent collision.

Victoria
In the Yukon, the Mackenzie Mountains, and Victoria Island, ~900 Ma rift-related clastics are overlain by ~820-780 Ma platformal carbonates. Extensional structures are present until about 815 Ma in the Yukon (19). For this segment, this may have been a successful rift from North China (20), and hence we take the rift-drift transition at 815 Ma, or North China may have been the conjugate to the western margin of Laurentia. The Victoria segment was intruded by the 719 Ma Franklin LIP (14) and then reactivated by latest Ediacaran to Cambrian rifting on both the northern and western margins. Because this collision was oblique and diachronous along the margin, we take the start date from the termination of collision on the Innuitian margin at 432 Ma to the end of the Silurian at 420 Ma. We separate this collision from the Ellesmerian orogeny, but acknowledge that these could be interpreted as a continuum with collision extending through the Devonian. As a strike-slip orogen, it is unclear if there was significant crustal material subducted to the mantle, and consequently we do not include this segment in calculations of subducted passive margin length.

North Slope
The North Slope terrane was displaced westward during the Paleozoic Innuitian and Ellesmerian orogenies (21), and is now in Arctic Alaska, where like the Victoria segment, 719 Ma plume-related magmatism was emplaced into a Tonian carbonate margin (22,23). This is followed by Cryogenian to Paleozoic glacial and carbonate deposition, which records a series of failed or outboard rifts until a successful rift around the Ediacaran-Cambrian boundary (22,24). In Arctic Alaska, passive margin termination is marked by a significant Upper Ordovician to Lower Devonian erosional unconformity, which we correlate with passive margin termination along the Innuitian margin (24).

4-5. Greenland & Svalbard
A vast thickness (12-15 km) of Tonian platformal strata define a rifted passive margin on the northeastern margin of Laurentia. In both northeast Svalbard and East Greenland, the Neoproterozoic successions begin with ca. 900 Ma rift-related clastic rocks, followed by Tonian platformal carbonates, ~1 km thick Cryogenian  successions, and thin early Ediacaran deposits (25,26). The rift-drift transition was estimated at 815 Ma (25,26). We define a diachronous onset of the Taconic-Caledonide orogenies between the Appalachians and the Arctic between 465 and 445 Ma, and take a 455 Ma onset age for the Greenland-Svalbard segment (27). We cut crustal contamination off at 443 Ma because Laurentia was the upper plate during the Caledonian orogeny with Baltica.

6-8. Appalachian
We divide the Appalachian margin into three segments that record the rifting of three separate cratons from Mesoproterozoic reconstructions of Rodina (28). The Scottish segment likely records the rifting of Baltica, the Northern Appalachian segment the rifting of Amazonia, and southern segment the rifting of Kalahari, and perhaps a later ribbon continent (29) such as Arequipa.
At least two pulses of plume-related intrusions of the Central Iapetan Magmatic Province (CIMP) were emplaced at ~615 and 590 Ma (30) between Laurentia, Baltica, and Amazonia. Rift-related volcanism occurred on the Appalachian margin of eastern North America between 562 and 550 Ma. The basal onlap on the distal cratonic margin are Middle Cambrian in age and we take a riftdrift age of 520 Ma for the Appalachian segment (31). Deposition in the Taconic forelands began by 465 Ma with arc-continent collision followed by slab breakoff and reversal (31), after which Laurentia was on the upper plate. We use these parameters for both the Scottish and northern Appalachian segments. For ease of visualization we place the Scottish segment on southern Greenland, but note that fragments are preserved in Scotland.
Tonian to Cryogenian rift-related magmatism dated at ~760-700 Ma is present on the southern Appalachian segment (32,33), but is absent north of the New York promontory. A Cryogenian to Ediacaran passive margin sequence is absent suggesting this event was locally a failed rift or farfield. A feasible scenario is that Tonian-Cryogenian rift related magmatism records the rifting of Kalahari, which was separated from North America by another continental terranes, perhaps the Arequipa terrane, which rifted away along with Amazonia in the latest Ediacaran to Cambrian. Consequently, we take the onset of rifting at 760 Ma, but take constraints on the successful rift and terminal collision from the northern Appalachian segment.
During the Alleghenian orogeny with Gondwana starting at ~320 Ma, the composite Laurentian margin was on the lower plate. We attribute rapid exhumation at ~295 Ma (34) to be associated with slab-breakoff.

Ouachita
Rift-related magmatism in New Mexico and Texas spans from ~539 to 508 Ma (35,36). The oldest platformal strata are latest Middle Cambrian, consistent with a rift-drift transition at ca. 500 Ma. Although the Ouachita-Alleghenian-Mauritanide belt does not preserve an ophiolite, vast sediment with volcanic debris was shed across North America and North Africa as well as tuffs found in these basins by ~320 Ma (37,38), which we take as the termination date of the passive margin. Laurentia is interpreted to have been on the lower plate of a continent-continent collision with South America (39); we extend crustal contamination of the mantle for 40 Myrs to 280 Ma.

10-11. Cordillera
The Cordilleran margin of western Laurentia formed through two episodes of rifting, a Tonian-Cryogenian failed rift, and a successful Ediacaran-Cambrian rift (40)(41)(42)(43). Extension, synsedimentary faulting, and rift-related volcanism began with the 777-719 Ma CHUMP (CHuar-Uinta Mountains-Pahrump) basins (44) and equivalents in Canada, which were deposited in narrow failed rifts (45). Rift-related volcanism persisted until 690 Ma (43,46), and syn-sedimentary faulting and unconformities persisted throughout the Cryogenian (47,48). There may have been a brief passive margin stage between 660 and 580 Ma recorded in the Mackenzie Mountain Supergroup(49), but the margin was reactivated at 570 Ma (50,51). What appears as passive margin sedimentation in the aftermath of the Marinoan glaciation may be continued activity masked by the profound post-Snowball transgressions at ca. 660 and 635 Ma (51). The margin was reactivated during the late Ediciacaran, as marked by additional unconformities, basement derived grits, and basaltic volcanism (40,52). The rift-drift transition was previously placed at the Precambrian-Cambrian boundary (41,42), but recent geochronology on the craton suggest broad subsidence did not occur until ca. 508 Ma (53).
In the northern Canadian Rockies and adjacent Alaska, Devonian siliciclastic rocks of the Imperial, Tuttle, and Nation River Formations, and the Earn Group represent a foreland basin (54), starting by 387 Ma (55). Terrane suturing continued through the Late Paleozoic, but may have been offboard (56). The southern Cordillera collided with an arc terrane during the Devonian Antler orogeny. Convergence began offshore in latest Devonian and platform drowning is Early Mississippian (57), with the death of the passive margin placed at 357 Ma (55).

Brookian
The Brooks Range marks a Mesozoic arc-continent collision between the Anguyuchum arc and the passive margin of the Arctic Alaska microcontinent, which was rifted from the Middle Devonian (~390 Ma) to earliest Carboniferous (58). The rift-drift transition is marked by platform carbonates of the Lisburne Group, which are as old as ~350 Ma (55,59). Arc-continent collision is marked by an influx of flysch from southerly sources, which began at 146 Ma (60). Exhumation ages from ca. 146-90 Ma are provided by deposition of foreland deposits on the North Slope that contain ophiolitic detritus (60). We extend crustal contamination in this oblique collision to 120 Ma to encompass shoaling in the foreland on the North Slope autochthon, which we associate with slab breakoff.

Scandanavia
Plume and rift related intrusions of the Central Iapetan Magmatic Province (CIMP) were emplaced between 616 Ma (30) and rift-related dikes at 608 Ma (61). We place the rift-drift transition at 605 Ma and the end of the passive margin at 505 Ma with arc-continent collision in the Finnmarkian orogeny (55,62). After arc-continent collision, the Baltican margin was subducted under composite Laurentia from ~435-415 Ma in the Caledonian orogeny (27). Thus, we mark crustal contamination from the subduction of Baltica between 505 and 415 Ma with a gap from 493-435 Ma.

14-15. Timanide
Mesoproterozoic to Tonian platformal carbonate and minor silciclastic rocks cover the East European Platform and are unconformably overlain by an Ediacaran siliciclastic sequence. The Mesoproterozoic and early Tonian units may represent an intercontinental basin or a rifted passive margin, but in either case it appears that the margin rifted again during the late Tonian, which was followed by late Tonian to Cryogenian passive margin deposition (63,64). Narrow Late Tonian rift basins formed in Sweeden, which accommodated the Vasingso Group and contain microfossils that have been correlated with 780-730 Ma assemblages in western North America (65). We interpret these interior basins as the manifestation of a successful rift and rift-drift transion at ~750 Ma.
A ~670 Ma ophiolite was obducted during the Timanide orogeny (66), which extended through the Ediacaran to ~530 Ma (67). We mark arc-continent collision by the appearance of 630-590 Ma detrital orthoclase in ca. 610-590 Ma strata that unconformably overlie Tonian units, and the presence of 609-571 Ma detrital phengite (67). Following slab-breakoff and reversal, the main phase of the Timanide orogeny occurred as an accretionary orogeny with Baltica in the upper plate. The Timanide orogeny and active margin continued through the Paleozoic (67).

Uralian 1&2
Upper Cambrian to Lower Ordovician rift facies date the onset of rifting of the Paleozoic Uralian margin to ~500 Ma (68), with a rift-drift transition at 477 Ma (55). The remnants of the Early Paleozoic subduction-accretion complexes occur in a belt between the East European and the West Siberian cratons (69). Magmatism and ophiolite generation in the Magnitogorsk arc and equivalents spans 488-392 Ma, with Baltica continental crust entering the subduction zone by 380 Ma and foreland deposition between 375-359 Ma during the early Uralian arc-continent collision (68). We extend arc-continental-terrane collision between Baltica, the Magnitogorsk arc, and Kazakhstania from 380-359 Ma. Additional subduction of the amalgamated Baltica and Kazakhstania likely occurred during the late Uralian orogeny with the final arrival of Siberia, with Siberia on the upper plate. Starting at ~300 Ma, Siberia collided with the eastern margin of the amalgamated arc terranes and Baltica forming the Permo-Carboniferous Uralian suture (70). We bracket continent-continent collision with deposition within the Uralian foreland basin from the Cisuralian to Capitanian (71) (300-260 Ma).

Tornquist
Half grabens imaged in Poland have been correlated to the basal Ediacaran stratigraphy in drill core (72). We take a rift age of 616-550 Ma after CIMP magmatism and the 551 Ma tuff at the top of the rift sequence to mark the rift-drift transition (72). An unconformity between the Middle Cambrian and the Ordovician is likely related to the Finnmarkian orogeny of the Scandanavian margin. Sedimentation continued through the Ordovician, with the Late Ordovician closure of the Tornquist Sea and collision of Avalonia along the Thor suture (73). Seismic data and the absence of any subduction related magmatism on the Baltica margin other than air-fall ash deposits suggests subduction towards the southwest (74). We place the end of the passive margin 450 Ma with Ordovician units succeeded by a Silurian foreland basin to 422 Ma, that was further metamorphosed by Scandian-Acadian deformation and later strike-slip motion (74).

18-20. Avalonia
Rifting of Avalonia began as a backarc rift during the Ediacaran at 595 Ma (75) and culminated with an early Cambrian cover sequence (76). The arrival of Avalonia to the Appalachian margin created the 421-400 Ma Acadian orogeny in North America and Europe, in which Avalonia was subducted under North America (77).

Variscan
Rifting of the Amorica spanned 419-407 Ma in the Rheno-Hercynian Zone with a passive margin end date at 347 Ma (55). This was followed by the collision of Gondwana, which created Hercynian-Variscan foreland basins beginning at 340 Ma in the Czech Republic (78) that remained active into the Westphalian (304 Ma) in South Wales (79) through the Stephanian (299 Ma) in Germany (80) and from the Middle Variscan through the Sephanian (~335-300 Ma) in Poland (81). Final collision and exhumation is marked by the emplacement of 330-300 Ma post-kinematic granites (82). We cut off the collision between Avalonia and Amorica at 330 Ma, and assign the later forelands and deformation to the collision with Gondwana.

Saxo-Thuringia
The Saxo-Thuringian margin of Armorica was part of Gondwana during the early Paleozoic and includes Cambrian conglomerates and Lower Ordovician mafic volcanic rocks (82,83). We pick an onset of rifting at 500 Ma with a passive margin defined from 444 Ma to 330 Ma. Collision in the Variscan orogen continued until ca. 300 Ma (82).

Alpine
Permian to Triassic rifting in western Europe was followed by a rift-drift transition at ~170 Ma with a passive margin duration from 170-43 Ma (55,84). Ophiolite generation in the Mesozoic to Cenozoic Alpine-Pontide belt occurred predominantly from 170-140 Ma (e.g. Betic, Chenaillet, Zermatt-Saas, External and Internal Ligurides, Calabrian, Corsica, Mirdita, and Pindos), during opening of the Alpine-Tethys Ocean (85). Subduction related metamorphism began in the Alpine Tethys by the Valangian  with uplift and erosion of the ophiolites above the Iberian plate and Eurasia primarily from 50-30 Ma (86), and we cut off passive margin subduction at the end of this interval.

NW Iberia
NW Iberia preserves a passive margin duration from 475-385 Ma (87), with the demise of the passive margin in the early stage of the Hercynian orogen. This margin may have been much more extensive but is largely overprinted by Alpine metamorphism. Exhumation extends to 365 Ma (87), which we attribute to slab-breakoff.

Greece Pindos
A passive margin formed on the Apulian microcontinent from 230-60 Ma (55), and we extend the onset of rifting to 250 Ma and collision from 60 to 40 Ma (88), as an early Alpine suture.

26-27. Isparta
A passive margin developed on the East Isparta margin from 227-53 Ma, and on the West Isparta segment from 227-60 Ma (55). We extend the onset of rifting to the Permian-Triassic boundary and end the collision at 35 and 40 Ma, respectively, with closure of the northern Sakarya zone, which formed a north-dipping Triassic subduction-accretion zone on the northern margin of the Paleo-Tethys (89), which was incorporated into the Pontides suture and uplifted from the Maastrichian through the Eocene (90). Since the Miocene, this suture zone was reactivated as part of the Inner Taurus and Zagros belts, and we continue passive margin subduction to the present.

N. Iran
The Paleo-Tethys opened during the Ordovician, subduction was initiated by the Devonian, and it closed from the Permian to Triassic with the diachronous collision of the Cimmerian ribbon continent in the Eo-Cimmerian orogeny. Middle Ordovician to Middle Devonian volcanic rocks have been attributed to rifting, with a 390 Ma rift-drift transition (55). An active margin developed on the southern Turan margin through the Permian with collisional arc deposits appearing on Iranian passive margin in the Triassic (95). Initial Eo-Cimmerian collision started at ~227 Ma, with slabbreakoff at ~200 Ma, and backarc rifting at ~180 Ma (95).

Himalayan I
The Lesser Himalaya preserves a Cryogenian rift followed by an Ediacaran to Cambrian passive margin sequence. We define a rift starting at 650 Ma and a passive margin from 635-502 Ma (55). Arc-continent collision terminated at 490 Ma prior to the development of an Ordovician active margin (96)

Himalayan II
For the younger Himalayan margin, rifting occurred between 330 and 271 Ma, and arc-continent collision began by 52 Ma along the Indus-Tsangpo and Shyok sutures (97), marked by the appearance of young volcanic material on the Indian margin (98). However, after the initial arccontinent collision, the subduction zone stepped north and consumed the Kshiroda plate (99,100) until collision between India and Eurasia starting at 40 Ma (97). We continue passive margin subduction through the continent-continent collision, which started at ~41 Ma (97) and continues to the present. Thus, we combine the subduction of Indian crust in two events starting with the arccontinent collision from 52-41 Ma and the continent-continent collision from 41 Ma to present.

32-33. Karakorum-Qiantang
The Karkorum block is the western extension of the Qiangtang terrane west of the Altyn-Tagh Fault. The West Jinsha suture is defined by the Triassic Yushu mélange between the Qiangtang terrane and the Songpan-Garze belt (101). A passive margin existed on the northern margin of the Qiangtang terrane from the Cambrian through the Permian, and was reactivated during the Permian with the rifting of a ribbon continent on the southern margin by the Early Permian (ca. 299 Ma). Magmatism within the oceanic tract is largely Permian in age (101). Subduction of the northern margin began by 230 Ma (101,102), after which an active continental margin was established by 210 Ma.

Taimyr
It was previously proposed that a Tonian ophiolite and arc collided with the Mesoproterozoic Taimyr margin at ~650 Ma (66,103). However, recent geochronology suggests that the Siberian margin was active after ca. 900 Ma and that ca. 730 Ma ophiolites formed in a back-arc and were accreted and exhumed during the Ediacaran (104); consequently we do not include the early Taimyr passive margin (55). After Ediacaran accretionary orogenesis the Taimyr margin was re-rifted in the late Ediacaran to early Cambrian with a rift-drift transition at ~525 Ma (105). A late Paleozoic orogeny was dated by Late Pennsylvanian to Early Permian thrusts and Permian granitic plutonism and metamorphism in the central Taimyr zone (106,107). This orogeny is considered a continuation of the Uralian suture. We date the onset of collision at 288 Ma with the age of the youngest suprasubduction granites (107). Ar-Ar ages of ~272 Ma represent termination of Late Paleozoic collisional tectonic activity within Northern Taimyr (107).

35-36. Yenisei
The Yenesei margin of Siberia rifted sometime after 1100 Ma to accommodate the Stenian to early Tonian Tungusik Group (108). Peak metamorphism of the margin occurred between 895-855 Ma (109). We define the passive margin from 1050-895 Ma. The margin then re-rifted between 800-790 Ma (110) with passive margin development by 715 Ma (109) to accommodate Cryogenian glacial deposits in the Chivda Formation (111). The margin then became active again with 630-610 Ma arc-continent collision with the Isakovaka arc (109,112).

Cis-Patom-Baikal
Cyrogenian magmatism associated with the Olokit rift occurred between 730-650 Ma (113), likely associated with separation from the northern margin of Laurentia. Passive margin deposits include Marinoan age ~645-635 Ma glacial deposits (114). The Neoproterozoic Baikal-Muya belt collided with the southern margin during the Edaicaran with foreland deposits that include the ca. 570 Ma Shuram excursion (114). We use the foreland to define passive margin subduction of the southern margin of the Siberian craton from 580-560 Ma. Cambrian to Ordovician oblique collisions along the southern margin of Siberia mark the subsequent accretion.

Verkhoyansk
Previous compilations defined three separate rifted passive margins on the eastern margin of Siberia (55). The first margin started at ca. 1600 Ma and ended at ca. 1010 Ma. This appears to be way too long-lived for a single passive margin and there is no record of a collision outside of what is interpreted as a foreland. The older succession could instead be part of the broad Mesoproterozoic intercratonic basins that formed on Siberia and Laurentia. If there is a Mesoproterozoic rifted passive margin, it is in the ca. 1100 Kerpyl Group, which lies unconformably over Lower Mesoproterozoic strata. Depending on the reconstruction, the eastern margin may have faced the Grenville orogen and all of the Late Mesoproterozoic units could be related to foreland deposition. Instead we interpret these successions to represent an intercontinental basin succeeded by distal foreland deposits of the Grenville.
Rifting of the eastern margin started at 543 Ma with a rift-drift transition at 523 Ma (115). This margin was reactivated with the emplacement of the Yakutsk LIP and separation of a ribbon continent at ~380 Ma and ended at ; because this margin was reactivated instead of terminated, we do not demarcate a separate passive margin death. We continue the Anguychum suture through to Chukotka, the South Anyui suture, and the Verkhoyansk of Russia, but because constraints are lacking in these belts, we largely use the parameters from the closure of the Anguychum Ocean. In the Verhoyansk, a Jurassic collision between the Mesozoic Alazeya arc and the Omelevka microcontinent is lined with ophiolites and created the Kolyma-Omion microncontinent, which was subsequently thrust over the Siberian margin (116).

Khubsugul-Zavkhan
We combine the Gargan margin with the Khubsugul and Zavkhan margins of Mongolia, which we interpret to have formed on the Tuva-Mongolia microcontinent that was exotic to Siberia until the Paleozoic (117). The passive margin formed during latest Tonian to Cryogenian rifting and ended with the collision of the Khantaishir-Agardhag arc between 545 and 525 Ma (117,118). Previous compilations also included the Idermeg terrane (119), but we find this margin too poorly constrained. Additionally, we interpret the Bayankhongor ophiolite as an oceanic plateau along an accretionary margin and do not use it as a constraint on a passive margin termination.

Tianshan-North Tarim
Volcanic rocks in the Quruqtagh Group on the northern margin of the Tarim craton have been dated between ~740 and 615 Ma and interpreted as a rifted passive margin (120), although others have interpreted the margin as a back-arc rift and long-lifted accretionary margin (121). Nonetheless, we place the rift-drift transition at 615 Ma and the initial collision at 455 Ma marked by an unconformity and influx of siliciclastic detritus (122). A 440-390 Ma belt of arc magmatism in the Tianshan and northern Tarim records the establishment of a south-dipping subduction zone under the Tarim by the Silurian, which was followed by Silurian-Devonian back-arc extension and accretion of the Yili block (123).

Kunlun
The Neoproterozoic stratigraphy on the southern margin of the Tarim block appears to mirror that on the northern margin, and consequently we follow constraints from the Quruqtagh for the rift and drift. Previous compilations defined the end of the passive margin at 430 Ma (55), but we instead suggest that this is a successor foreland basin associated with the collision of the and that the peripheral foreland associated with collision of the South West Kunlun arc terrane formed along the Kudi-Altyn suture from 475-455 Ma (124). We assign the later collision to the subduction of composite Qilian-Qaidam-North Qinling and North China block below the Tarim block (125).

North China
Tonian basins were deposited across North China (126) above the ~970-890 Ma Xuhuai rift system. These basins were re-activated during the Cambrian and succeeded by Late Ordovician foreland deposits (127). We define the passive margin from 890-455 Ma, and termination with the collision of the Erlangping ophiolite. The Erlangping suture between an Early Ordovician oceanic arc and the North China craton, slightly proceeds the Qilian suture, with collision over by 435 Ma (128). We place the subduction of North China below these terranes from 455-435 Ma.

43-44. Longmen Shan and Qinling-Dabie
After collision and accretion of the Yangtze and Cathaysia blocks by 810 Ma, the northwestern margin of the South China craton became an active margin through much of the Cryogenian. By the latest Cryogenian, the southern margin of the Yangtze block had developed into a rifted passive margin, perhaps through back arc rifting, however, we find the Ediacaran northern margin of South China too poorly constrained to define as a passive margin (55). Active Tonian to Cryogenian arc magmatism on the northwest margin suggests the margin may instead have been formed by backarc extension. We attribute Cambrian-Ordovician strata to the southern margin, and define Silurian rifting for the northern margin (129), and a passive margin by 400 Ma. The margin was reactivated at 300 Ma (55), and the passive margin ended with the Dabie-Sulu orogen starting at 228-210 Ma. The Dabie-Sulu suture was exhumed in the late Triassic to Jurassic (228- 210 Ma) during the final collision between North China and South China (130), and extends east to Korea (131).

Nanling
Rifting on the southern margin of the Yangtze block occurred during the Sturtian glaciation culminating with horst and graben structure capped by the post-Sturtian transgression. Thus, we define rifting from 720-660 Ma. Termination of the passive margin occurred by 450 Ma with putative distal retro-arc foreland basin deposition associated with the Kwangsian orogeny (132,133). However, it appears this margin was never subducted but instead reactivated in a back arc to accretionary setting after establishment of east dipping subduction outboard, and consequently, we do not include the Nanling segment in the calculation of subducted passive margin length.

Taiwan
A passive margin developed on Taiwan from . Taiwan is one of the best-constrained examples of active arc-continent collision and subduction polarity reversal. Collision began at 6 Ma and exhumation has accelerated over the past million years (134).

Timor
Rifting had started on the northwest margin of Australia by the latest Triassic with a rift-drift transition at 151 Ma (55). The passive margin end date is about 4 Ma and the collision of the Banda arc continues today (135).

New Guinea
Arc-continent collision began in New Guinea during the Late Oligocene to Early Miocene above a north-dipping slab (136)(137)(138). Two major ophiolite belts-the Irian-Marum ophiolite belts (including the April ultramafics), and the Papuan Ultramafic Belt (PUB)-are preserved along the Central and Peninsular Range. Exhumation of the Irian ophiolite began in the middle Miocene (16- 14 Ma), and uplift in the Central Range accelerated from the Late Miocene to Pliocene (139). Although the PUB was generated and obducted earlier than the ophiolites in the Central Range (140,141), it was also exhumed very rapidly over the past 10 Ma (137).

49-50. Centralian & Adelaide
When Australia rifted from Laurentia during the Tonian, basins formed throughout Australia, and a rifted passive margin developed on the eastern margin (142). We include the Kimberley with basins of Central Australia. Rift-related magmatism has been dated between 825-750 Ma, and we pick the rift-drift transition at 750 Ma (143). Collision between North and South Australia in the Paterson-Peterman orogeny created large-scale metamorphism and foreland deposition between ~570-530 Ma (144), which was manifested in the influx of siliciclastic material in the Georgina Basin (142). Peak metamorphism occurred between 550-530 Ma above a south-dipping slab (145). On the Adelaide margin, we associate canyon cutting in the ~570 Ma Wonoka Formation to be associated with disruption of the passive margin, followed by an influx of clastic material in the Pound Group from developing highlands to the northwest (146). However, it appears this margin was never subducted but instead reactivated in a back arc to accretionary setting after establishment of east dipping subduction outboard in the Ross-Delamerian orogeny (147). Consequently, we do not include the Adelaide segment in the calculation of subducted passive margin length.

Tasman
The western rifted margin of Australia continues to Tasmania and hosts Tonian to Cryogenian deposits of the Black River Group (148). Volcanic rocks of the Rocky Cape Group and equivalents formed between ~582-575 Ma (149), which we associate with back-arc rifting and consequently we cut passive margin sedimentation off earlier in this segment at 600 Ma. In the Ross Orogen of Antarctica, an active arc was established by 565 Ma (150). In Australia, the Delamerian orogeny formed from 520-490 Ma due to accretion of an outboard arc (147), with peak metamorphism in Tasmania between 520 and 508 Ma (151). Like the Adelaide segment, there is no evidence for subduction of the passive margin, so we do not include this in the calculation.

Cuba
Cuba hosts a carbonate-dominated Mesozoic sequence that formed during the break-up of Pangea with a rift-drift transition at ca. 159 Ma (55). Volcaniclastics from the Greater Antilles arc appeared on the margin at ca. 80 Ma. We extend the rift to 200 Ma to encompass Jurassic volcanism and subduction of the margin until 60 Ma (152).

Venezuela
Rifting of Pangea began with the ca. 200 Ma Central Atlantic Magmatic Province, and we follow Cuba for the rift-drift transition at ca. 159 Ma (55). The transition from passive margin to a foredeep is defined by an increase in subsidence at ~34 Ma (153), and passive margin subduction until 8 Ma.

Araras, Paraguay
Rifting occurred during the Marinoan glaciation, forming grabens filled with glacial deposits and iron formation. We define rifting from 645-635 Ma. The SE margin of the Amazon craton is marked by the 540-500 Ma Paraguay and Araguaia belts (154), which are separated from the Goias magmatic arc by the Transbrasiliano Lineament. This marks the continent-continent collision between the Amazon, West Africa, and Sao Francisco cratons. The SE margin of Amazonia has alternatively been interpreted as an active continental arc throughout most of the Neoproterozoic (155), however, the Goais arc formed as an inter-oceanic arc that collided with the Sao Francisco craton and there is no evidence for subduction under Amazonia at this time (156). The continental arc likely formed on the Sao Francisco craton from ~630-560 Ma, after ~650-630 Ma collision with the Goias arc.
A Cryogenian rifted passive margin deposit with a Marinoan cap carbonate is present in the Paraguyay Belt (157). This early Ediacaran succession is unconformably overlain with 550-540 Ma foredeep deposits of the Tamengo and Buaicurus formations (158) and early Cambrian foreland deposits of the Diamantino Formation (159). Deformation and metamorphism is bracketed by 518 Ma undeformed granite (159).

Iapetan margin of Amazonia
The Iapetan margin of Amazonia is poorly represented in heavily deformed para-autchthonous belts of the Maranon complex, but is preserved in part in Late Ediacaran to Cambrian sequences that unconformably overlie basement and Cryogenian strata near the Bolivia-Brazil border. We pick CIMP rifting from 616 to 570 Ma (30). The margin ended with the Ediacaran-Late Cambrian oblique collision of the Pampean terranes and Arequipa from 545-520 Ma (160).

Arequipa
Arequipa formed as a ribbon continent between Laurentia, Amazonia, and the Kalahari craton during the Cryogenian (161). Unconformities developed until an Ediacaran carbonate platform blanketed the outcrop belt. We place the rift-drift transition at ~630 Ma. The margin was drowned by siliciclastic sedimentation between 545 and 520 Ma (160).

Precordillera
Rifting between the Precordillera and the southern margin of North America occurred between ~539-508 Ma (35,36). The oldest platformal strata are latest Middle Cambrian, consistent with a rift-drift transition at ca. 500 Ma. Foreland deposition began in the Early Ordovician and continued through the Ordovician (162).

West Sao Francisco
The Brasiliano orogeny involved a collision between the Bambui platform on the west side of the Sao Francisco craton, and terranes to the west. Extension-related magmatism in the Sao Francisco craton has been dated at 906 Ma and between 800-760 Ma, and interpreted as a back-arc rift of the Goais block from Sao Francisco (163). These are overlain with a passive margin sequence that includes Sturtian age glacial deposits (164). The Brasiliano orogeny involved a collision between the Bambui platform on the west side of the Sao Francisco craton, and terranes to the south and west including the Goais arc and the Paranapanema block. Magmatic ages in the Goias arc come in two major pulses, one from ~890-860 Ma and a second between ~670-600 Ma (165). U-Pb zircon dates on metamorphic overgrowths constrain high-grade metamorphism along the eastern boundary of the Goais massif at 760-740 Ma (166). Peak metamorphism in the Brasiliano belt was between 650-610 Ma (167). We interpret 760-740 Ma metamorphism to record arc-terrane collision off-board, which was followed by 630-610 Ma continent-continent collision with the Paranapanema block and termination of the passive margin.

East Sao Francisco
For rifting, we follow constrains from West Sao Francisco. The Socorro and Serra da Bolívia magmatic arcs collided with the already amalgamated Paranapanema-São Francisco plates between 620 and 605 Ma (168). This was followed by accretion of the Rio Negro magmatic arc of the Oriental terrane between 605 and 595 Ma. Widespread generation of crustal melts associated with collision is represented by foliated granitic plutons, dated between 610-565 Ma (168). Although collision of the Riberia and Dom Feliciano belts certainly overlapped, we associate collision of the Socorro and Serra da Bolívia magmatic arcs with the Dom Feliciano suture, and take the Ribiera suture from 610-590 Ma. This was followed by oblique continent-continent collision with Congo through the early Cambrian in which Sao Francisco and Rio de la Plata were on the lower plate (169), and consequently we extend subduction of the margin until 540 Ma.

60-62. Sierra de la Ventana, Cape Fold Belt, Ellsworth Mountains
In the Cape belt, Middle Cambrian rift deposits are overlain by an Early Ordovician to Carboniferous passive margin sequence (170). Foreland basin deposits of the Karoo Group formed between 300-280 Ma (171). This margin has been correlated with equivalent units in the Sierra de la Ventana belt in Argentina and the Ellsworth Mountains of Argentina (55).

East margin of West African craton
A Neoproterozoic rifted passive margin developed on the eastern margin of the West African craton between 1000-635 Ma, which preserves a ca. 635 Ma basal Ediacaran cap carbonate in the Volta Basin (172). Neoproterozoic ophiolites of the Buem belt were exhumed after ~710 Ma in an arccontinent collision (173,174) associated with 620-602 Ma UHP metamorphism (175). We take these ages as the best constraint on exhumation during arc-continent collision in the Dahomeyide belt. These are overprint by ~601-570 Ma metamorphism related to transpressional contintentcontinent collision with the Nigerian Shield (175).
The Trans-Saharan continent-continent collision between the West African craton and the Tuareg-Nigerian shield occurred between 601-567 Ma, as marked by migmitization and foreland basin development (175). A maximum age on foreland deposition comes from a 601 Ma tuff directly below the main foreland basin succession in the Oti-Pendjari Group of the Volta Basin (176). A 601-567 Ma collision is further consistent with 586-567 Ma Ar-Ar and titanite ages from within the Dahomeyide thrust stack (177).

North margin of West African craton
In the Anti-Atlas belt, on the West African Craton, south of the Anti-Atlas Major Fault, 2200-2030 Ma basement is overlain by volcaniclastic rocks of the Taghdout Group and Tachdamt Formation, which form the lower portion of the Anti-Atlas Supergroup and have been dated at ~883 Ma (178,179). Volcanic units have sub-alkaline to tholeiitic geochemical signatures, which have been interpreted to record the development of a volcanic rifted passive margin (180). North of the Anti-Atlas Major Fault, the Sirwa and Bou Azzer ophiolites have been dated between 762-759 Ma (181,182). Crustal thickening and a magmatic lull from 730-710 Ma has been related to arc-continent collision between the Sirwa and Bou Azzer ophiolites and the West African craton above a northdipping subduction zone (183). After collision, an active continental arc was established in Morocco by 710 Ma (183) and on crustal fragments of Cadomia and West Avalonia through the Early Ediacaran, followed by Late Ediacaran backarc rifting of the Cadomian arc (184).

West margin of West African craton
A late Mesoproterozoic rift accommodated the ~1100 Ma Atar Group in Mauritania (185). The margin was reactivated with a thick Tonian siliciclastic sequence in the Assabet Group, which coincided with rifting of the northern margin and potentially collision on the southwestern margin (186). We follow constraints from the northwest margin of West Africa and propose that the margin terminated with an arc-continent collision between 730-710 Ma. In allochthonous units, initiation of subduction is defined by the ~710 Ma Gorgol Noir ophiolite (187), which we associate with the transformation of the margin to an active arc. The Bassaride-Rokelide belt of Guinea, Sengal, and Sierra Leon, extends into the Souttouf belt of Western Sahara where there is extensive 660-650 Ma metamorphism (188,189). In Mauritania, an unconformity at this level is overlain by 640-635 Ma Marinoan glacial deposits (190). In the Souttouf belt, metamorphic ages associated with accretion cluster between 610-590 Ma (191).

66-67. Northwest Congo-North Sao Francisco
Neoproterozoic metasediments and granites are preserved on the northwestern margin of the Congo Craton in the Central African Fold Belt. Granites have been dated between 641-613 Ma and metamorphism between 620-610 Ma marks the demise of the margin (192). The margin extends to the NW margin of the Sao Francisco craton and was reactivated as a dextral transcurrent margin at ca. 570 Ma. On the Sao Francisco craton, rifting occurred at 806 Ma (193) and accommodated deposition of Cryogenian strata of the Vaza Barris Group (194). A Tonian rift basin is present in the Lower Dja Series in Cameroon, which is overlain by Cryogenian passive margin deposits of the Mintom Formation (195). We take rifting at 806-750 Ma, passive margin deposition to 620 Ma, and subduction of the passive margin to 600 Ma.
On the northern margin of the Sao Francisco craton the ~820 Ma Monte Orebe ophiolite is associated with an ocean-continent transition (196). The passive margin includes both Cryogenian glacial deposits and cap carbonates (197), suggesting deposition until at least 635 Ma. An external arc was active from 650-610 Ma (198) with syn-orogenic metamorphism between 610-595 Ma. We take foreland deposition related to continent-continent collision from 615-595 Ma. The Sergipano belt extends to the Rio Preto belt on the NW margin on the Sao Francisco craton and the Oubanguides in Cameroon (199).

68-69. Southwest margin of the Congo
The western margin of the Otavi formed rifted between 770 and 655 Ma, and terminated with Ediacaran collision of the Outjo block and foreland deposition (200). In the Coastal Terrane, peak metamorphism occurred from 650-640 Ma (201), which we suggest is associated with relict subuduction and pre-collisional. Collision is dated by ca. 580-570 Ma syn-kinematic metamorphic granites and 590-570 Ma molasse of the Mulden Group. After the main phase of collision, transcurrent slip and erosion continued from 570-530 Ma, which was followed by rapid exhumation and thus presumably erosion during transtensional reactivation 525-520 Ma (201).

Zambezi-Mpanshya
Rift-related magmatism in the Zambezi belt extended from 804-735 Ma, with a passive margin from 735-585 Ma (202). Eclogite in the West Zambezi Belt indicates that subduction was underway by ~659-638 Ma (203). Metamorphic ages constrain ocean basin closure and collision of the Congo and Kalahari cratons by ~585-565. Collision within the Zambezi Belt was ~10-30 Myr before collision in the Damara Belt at ~555-550 Ma (202).

Karasuk
In Uganda, a passive margin is preserved in the Karasuk Supergroup with less deformed equivalent strata inboard in the Malagarasi Supergroup, which include Cryogenian glacial deposits of the Bunyoro Group (204). The West Granulite belt, is a continuation of the Malawi suture between the amalgamated ANS terrante and the Tanzania/Congo craton (202), and is marked by the Sekker and Moroto ophiolites. The West Granulite belt records two pulses of high PT metamorphism at ~657-639 Ma and ~635-615 Ma followed by 575-525 Ma sinistral transpression (202). We interpret the 657-639 Ma dates to mark exhumation on the West Granulite suture and the 635-615 Ma dates to record final amalgamation and collision in the East Granulite belt.

72-73. Northwest and northeast margins of the Kalahri
Like the southern Congo margin, rift-related magmatism was widespread after ~805 Ma (202), with passive margin sequences deposited between ~720-570 Ma on the Kalahari craton. On the northwest margin, collision occurred between ~570-515 Ma as defined by the youngest ages from the basal Sijarira Group (202). After collision of the Zimbabwe promontory, the Kalahari craton rotated clockwise and closed the Khomas Ocean during the 555-515 Ma Damara Orogen (202). We define the passive margin termination with deposition of the ~555-535 Ma Nama foreland basins, with metamorphism continuing through ~515 Ma with Kalahari on the lower plate, subducting below the Congo. Figure S1. The distribution of Eu anomalies for a compilation of upper continental crust rocks is compared to the Eu anomalies for lower continental crust rocks. A database (205,206) of granulites (representing lower continental crust) and glacial tillites, loess, greywacke and shale (representing upper continental crust) shows that both continental crustal types exhibit considerable overlap in Eu/Eu* values (Eu/Eu* = EuN/(SmN * GdN) 0.5 , where N represents normalization to primitive mantle from elsewhere (207)), and both type of continental crust exhibit high Eu/Eu* (>1) and low Eu/Eu* (<1) values. Low Eu/Eu* values (i.e., Eu/Eu* < 1) from subducted upper continental crust have been argued to be present in the mantle sources of EM2 OIB, and high Eu/Eu* values (i.e., Eu/Eu* > 1) from lower continental crust were argued to be in EM1 OIB (208), but the overlap in Eu/Eu* between upper and lower continental crust, and Eu/Eu* > 1 in examples from both upper and lower continental crustal rocks, indicate that Eu/Eu* cannot be used to distinguish between recycled upper and lower continental crust contributions in OIB. Note that the distribution of lower continental crust rocks exhibits a long tail of infrequent high Eu/Eu* values that extend off the range shown in the figure.  Dataset S1 (separate file). Data for 58 known hotspots summarized from Jackson et al. (1). The data include surface locations of hotspots, locations for the bases of advected plume conduits at 2850 km (where locations represent averages across four models), average plume conduit distances from the LLSVPs at 2850 km, and hotspot buoyancy fluxes.
Dataset S2 (separate file). Area normalized paleolatitude from 1000-520 Ma. Continental areas were calculated in QGIS, including each of the major cratonic blocks larger than Nigeria-Benin; smaller continents were not included because of uncertainties in their size and paleolatitude (see Methods text). Latitude was extracted from the latitudinal midpoint at 5º at 20 Myr intervals from Merdith et al. (210), with modifications following Eyster et al. (212). See Torsvik et al. (213) for 540-0 Ma.
Dataset S3 (separate file). Passive margins of the past billion years, described in SI Appendix text by ID number with references. Passive end is equivalent to collision start. Continent and Plate ID are used for paleogeographic reconstruction in GPlates. Facing direction is the direction of passive margin subduction at the beginning of the orogen. Paleolatitude of the margin at the passive margin end date is to the nearest 5 degrees. Data are used to calculate passive margin lengths in Dataset S5.
Dataset S4 (five separate files). Five separate files for use in GPlates software. These files provide the shapefiles for passive margins. The shapefiles are used to generate passive margins that are compiled in Datasets S3 and S5.
Dataset S5 (separate file). Summed length of passive margins and passive margin terminations. Passive margins are calculated in 5 Ma intervals between start and end date. Passive margin terminations are calculated in 5 Ma intervals between passive margin end and collision end date, from Dataset S3 and the SI Appendix. Length in southern or northern hemisphere designates the hemisphere in paleogeographic reconstruction at the time of passive margin end date.
Movie S1 (separate file). Provided in .mp4 format. The movie shows passive margin terminations over the past 1 billion years. Passive margins from Datasets S3 and S5 are shown on a plate reconstruction in 5 Myr frames. Passive margin shapefiles used in the Supplementary Video are available in the Dataset S4. The Dupal formation interval (650-300 Ma) is indicated.