A tectonic-rules-based mantle reference frame since 1 billion years ago – implications for supercontinent cycles and plate–mantle system evolution

For most of geology's history, the deep mantle was a black box. Geologists could track how plates moved relative to each other — Africa pulling away from South America, the Pacific opening and closing — but knowing where those plates actually sat on the globe, anchored to the convecting mantle beneath them, was a different problem entirely. It remained unsolved because every reference frame researchers tried eventually broke down. Hotspot tracks work beautifully for the last one hundred million years, but preserved, dated chains become sparse quickly as you go deeper into time. Paleomagnetism gives you latitude and orientation, but not longitude. The Earth's magnetic field is radially symmetric, so it simply cannot tell you where a plate sat from east to west. And anchoring reconstructions to the large low-shear-velocity provinces, or LLSVPs, those vast slow anomalies sitting at the base of the mantle — that idea seems attractive until you look closely and find that statistical tests can't confirm their edges are where plumes rise, and flow models suggest their shapes are controlled by subducting slabs rather than being primordial fixtures. Multiple studies concluded that the African LLSVP in its present form probably didn't exist before about 230 million years ago. So hotspots fail in deep time, paleomagnetism is longitudinally blind, and deep-mantle anchors may not be stable enough to anchor anything. That impasse is where Müller and colleagues stepped in, and their solution is conceptually elegant. Instead of anchoring the reference frame to a physical feature that might itself have moved, they asked a different question: what does well-behaved plate tectonics actually look like, and can we find the reference frame that best satisfies those rules? The result is SEEM1000 — the Solid Earth Evolution Model for the last one billion years. The team defined four tectonic constraints and jointly optimized a plate motion frame to minimize their combined misfit. Net lithospheric rotation — the average rotation of all plates together — should be small. Trench migration rates should be bounded. Continental velocities shouldn't be geologically absurd. And for the last eighty million years, where the data exist, the model should match age-progressive hotspot tracks. The optimization runs as an iterative, seed-based inversion, stepping backwards through time in five million year increments. At each step, four hundred candidate reference rotations are generated and locally optimized in parallel, and the global minimum is selected. The weighted objective function includes explicit penalties. If net rotation climbs above 0.20 degrees per million years, or if median trench-orthogonal velocity exceeds thirty millimeters per year, the solution is penalized. What comes out is a reference frame where net lithospheric rotation stays consistently below 0.25 degrees per million years, in agreement with independent geodynamic models. There is no single physical anchor. Just physics-based rules, applied globally, across a billion years. Once the reference frame is in hand, the first major result it delivers is a temporal contrast that turns out to be dramatic. Müller and colleagues nicknamed the interval from 600 to 320 million years ago the "zippy tricentenary," and the name earns its keep. During this window, trench motion scatter is roughly double what it is in any comparable recent interval. The bulk of trench migration rates ranges between three and six centimeters per year, with larger outliers driven by a predominance of short-lived, highly mobile subduction zones. Multiple internal ocean basins were opening and closing rapidly as Gondwana assembled and the Iapetus and Rheic ocean systems evolved. Subduction was fast, spatially fragmented, and geologically brief. Contrast that with the last three hundred twenty million years — the Pangea assembly, stability, and dispersal interval. Trench motions are confined to a narrow band of roughly minus two to plus two centimeters per year. The median absolute deviation of subduction zone migration is largely between zero and one centimeter per year for trench advance, with slightly more for retreat. Even a pulse in the Early Cretaceous only pushes the scatter up to about four centimeters per year before it settles back down. A large, stable continental mass ringed by relatively immobile subduction zones is a fundamentally different dynamic regime from the zippy tricentenary's swarm of short, rapidly evolving trenches. This isn't a marginal statistical difference; it's a factor of two in scatter, marking what looks like a distinct chapter of Earth's tectonic behavior. The reference frame also speaks to one of the oldest debates in supercontinent science: the orthoversion hypothesis. The idea, first formalized by Mitchell and colleagues from paleomagnetic data, is that successive supercontinents form roughly ninety degrees of longitude apart, above subduction girdles that are orthogonal to where the previous supercontinent sat. Mitchell's group favored a westward offset of Pangea relative to Rodinia, partly because it minimized inferred plate speeds. Müller and colleagues recover an orthoversion-consistent outcome by an entirely independent route. Their optimized reconstruction implies Pangea assembled about ninety degrees east of Rodinia — the opposite direction. The eastward solution emerges naturally from the optimization constraints without being imposed on the model. Because Mitchell's original paleomagnetic approach couldn't discriminate between eastward and westward rotation — the magnetic record is longitudinally blind — the two approaches genuinely disagree on the geometry in a way that has testable consequences. If the continents moved eastward after Rodinia's breakup, eastern continental margins should have been flooded first as they encountered subduction-driven topographic lows. A westward migration predicts the opposite. That is a prediction geologists can go look for in the stratigraphic record. Now go deeper — literally. The plate motion model is coupled to mantle flow simulations, and together they reveal five distinct chapters of deep mantle evolution across the billion-year span. From one thousand to six hundred million years ago, the lowermost mantle hosts a broad, heterogeneous network of basal mantle ridges — no coherent large upwelling, just a tangle of smaller structures reflecting widely distributed, rapidly evolving subduction above. Between six hundred and five hundred million years ago, something more organized briefly appears: a transient degree-two configuration, meaning two roughly antipodal upwelling regions centered near the poles, driven by a band of low-latitude subduction. A degree-two planform consists of essentially two hemispheres of rising hot material facing each other across the planet. It's striking, and it doesn't last. From five hundred to four hundred million years ago, those structures migrate and fall apart. The northern basal structure drifts southward and begins evolving into what will become a Pacific-centered upwelling. The southern structure is cut apart by newly arriving subducting slabs and dissolves back into a network of ridges. The model shows why this takes so long: it requires roughly two hundred million years for a region of the lower mantle to register the absence of subduction above it, and about one hundred fifty million years for a coherent hot basal structure to assemble once that absence is established. Mantle memory is slow. From four hundred to two hundred million years ago, a stable degree-one planform emerges — one dominant upwelling hemisphere, centered on the Pacific. No symmetric antipodal counterpart develops because the African hemisphere remains populated by slabs descending from the closure of the Iapetus and Rheic oceans and subduction around Laurentia, Baltica, and Siberia. One hemisphere rising, one hemisphere sinking. After Pangea's breakup, as the African hemisphere's slab burden eventually fades, a basal degree-two structure re-emerges — foreshadowing the configuration we imagine with seismic tomography today. The model's fidelity to what we actually observe is quantified through comparison with seven independent seismic tomographic models. The preferred solution, OPT1, reproduces large-scale lower-mantle structure with a mean accuracy of seventy-two percent and a mean sensitivity of sixty-one percent averaged across those seven models. That's not perfection, but for a model running a billion years with no free parameters tuned to match tomography, it is meaningful validation. What Müller and colleagues have built is the first physically self-consistent reference frame covering a full billion years, coupled to mantle dynamics and validated against the seismic record of the present-day mantle. It answers questions that previously had to be approached with separate, incompatible tools: how large low-shear-velocity provinces form and how long they persist, how supercontinents organize mantle convection, and how heat escapes from the core over geological timescales. The code and data are publicly available, making SEEM1000 a living platform other teams can extend, test, and challenge. The deep finding — if you'll allow it — is that Earth's surface and its deep interior have been in continuous conversation for a billion years, and we now have a framework coherent enough to listen in. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.