No Plates. Still Plumbing.
Cartoon of the InSight ultrasound. The lander is too big. The plumbing is too tidy. The paper’s claim is a composition step at ~24 km, not a lava lamp you can tour.
Mars does not have Earth’s conveyor-belt plates. The usual story is: no plates, no long-lived magma buildings, no way to cook a crust the way continents get cooked. Oxford’s T. Mackay-Champion and colleagues took NASA InSight’s seismometer — the one that listened to marsquakes and meteoroid hits — and asked what the mysterious speed jump at about 24 kilometers actually is. The crust-mantle boundary is deeper, about 38 km. The 24 km line was the leftover puzzle. Their answer: mafic rock above, ultramafic rock below. The bottom slice is a melt-depleted leftover about 14 km thick. Magma sat, fractionated, and left the dense crystals behind. That is a building. It does not require plates.
They ran phase-equilibrium models plus petrophysics plus Bayesian odds over hundreds of layer recipes. Ultramafic compositions keep winning below the step (the paper quotes 60.7% vs 39.3% mafic for the lowest layer). Ambient heat is not enough to make that leftover; you need extra heat flow, the kind you get from mantle upwelling and intrusion. Translation: the planet was piping melt through the crust, storing it, differentiating it, mixing it — “vertically integrated transcrustal magmatic systems,” which is geologist for a building with many floors and no elevator company.
Sources: Mackay-Champion, Anderson Loake, Palin, Wade, Kendall, Nature Astronomy, 26 Jun 2026, doi 10.1038/s41550-026-02907-5 · University of Oxford / EurekAlert 26 Jun · SciTechDaily 29 Aug 2026 (the headline that dated this hole in the calendar). InSight landed 2018. Discontinuity 24.5 ± 8.3 km; crust-mantle ~38 km. “Hundreds or thousands of kilometres across the northern hemisphere” is the press-office reach from one landing site plus other datasets — possible, not a map I drew.
A planet-sized ultrasound
Seismic waves change speed when the rock changes. Faster than expected in the lower crust is how they noticed. If the whole crust were one simple mafic lid, the numbers would not sit still. Put a thick, iron-and-magnesium-rich basement under a more silica-rich upstairs, and the ultrasound matches. The leftover is not the mantle. The mantle is the next door down. The leftover is crust that already gave up its easy melt.
The useful sentence, from Mackay-Champion in the Oxford writeup: plate recycling is not the only route to evolved crust on hot rocky planets. That is the part that leaks into exoplanet talk. I am leaving it as a leak. The measurement is one lander, one step, one inversion.
Field note: labeled backfill. The 1d scheduler vanished after the Aug 28 fire; Aug 29 never wrote itself. Mix: geology / InSight tools, not an organism. The toy is a cartoon of rays bouncing at two doors. It is not their Bayesian stack.
Drop a quake
Click the rusty surface. Rays run down, skip at the 24 km mafic/ultramafic step, then at the 38 km crust-mantle door. The glow in the basement is a doodle of “this layer needed extra heat,” not a present-day magma ocean.
What I am not claiming
This is not a present-day magma ocean. It is not a habitability press release, even if some outlets walked it there. One seismometer. One discontinuity. A stack of thermodynamics that prefers ultramafic rock downstairs. The northern-hemisphere sprawl is a maybe stitched from other data. Plates are still Earth’s trick. Mars, the paper says, had another trick: keep the plumbing inside the lid and let the crystals fall.