The mantle doesn't break. That's the textbook claim — below the Mohorovičić discontinuity, rock is too hot, too ductile, too plastic to fracture. Earthquakes happen in the crust. The mantle flows.
Shiqi Wang and Simon Klemperer at Stanford just catalogued 459 earthquakes in the continental mantle since 1990. Not subduction zones, where cold oceanic crust plunging into the mantle can snap — continental mantle, the warm semisolid stuff directly beneath our feet. Forty-six thousand candidate earthquakes, filtered by a technique so clean it makes the previous decades of debate look like guesswork.
The technique: Sn/Lg wave ratio discrimination. Sn waves travel through the uppermost mantle. Lg waves propagate through the crust. An earthquake in the crust produces strong Lg relative to Sn. An earthquake in the mantle produces the opposite. You don't need to know the exact depth. You don't need precise crustal thickness models. You just compare what comes through which medium.
The old approach was to estimate depth from arrival times, then subtract the local crustal thickness. But crustal thickness varies — the Tibetan Plateau has 70 km of crust where most continents have 35. A depth estimate of 45 km is a mantle earthquake in Kansas and a mid-crustal event in Tibet. Depth estimation compounds uncertainties. Sn/Lg ratios sidestep them entirely. The signal tells you which floor the rupture happened on, regardless of how tall the building is.
Four hundred and fifty-nine events cluster in two regions: under the Himalayas, where the Indian plate drives into Asia, and near the Bering Strait, where North American and Eurasian plates interact. The clustering isn't random — it follows the geometry of continental collision. The stress doesn't stop at the Moho.
This is the part that rewrites the textbook. The brittle-ductile transition was treated as a floor — below it, rock deforms plastically, no fracture. But Wang and Klemperer's temperature estimates from several regions exceed the hypothesized brittle failure threshold. The mantle isn't failing because it's cold enough to be brittle. It's failing while warm. The threshold itself was wrong, or the threshold model is incomplete, or brittle and ductile aren't the binary the textbooks drew.
“Continental mantle earthquakes might be part of an inherently interconnected earthquake cycle,” Wang says, “both from the crust and also the upper mantle.” The crust breaks; the mantle breaks; the distinction between where-it-breaks and where-it-flows was always a gradient, not a boundary. Four hundred and fifty-nine data points say so.
The diagnostic was there all along. Every one of those earthquakes was recorded by global seismic networks. The waveforms were sitting in databases. What changed wasn't the instruments or the data — it was the question. Instead of asking “how deep was this earthquake?” and then arguing about crustal thickness, Wang and Klemperer asked “which medium did the signal travel through?” The answer was already encoded in the relative amplitude of two wave types. It just needed someone to listen to both at once.