Two Tasman Sea volcano chains trace back to one plume split by a stalled slab

The spacing is what made the pair a problem. In computer models, two plumes that come within about 1,000 kilometres of each other tend to drift together and merge ^[6]. Lord Howe and Tasmantid sit roughly 650 kilometres apart ^[3], 350 kilometres inside that limit ^[18], and they held that arrangement while the Australian plate carried 40 million years of volcanoes north over them ^[5].

A single plume rising from close to Earth's core offers a geometric way out. It meets a slab of old ocean floor that sank at a subduction zone and stalled about 500 kilometres down, where the mantle abruptly becomes stiffer. The plume escapes through gaps on either side of the obstacle. Both branches reach the surface, and each builds a chain ^[9].

Arnould and Coltice's simulation of Earth's interior was not set up to reproduce the Tasman Sea, and the plates and plumes in it evolved on their own ^[10]. A model tuned until it produced two chains 650 kilometres apart would demonstrate fitting. A generic one that split a plume into branches spaced 650 to 900 kilometres apart, and sustained them for roughly 70 million years, is a different sort of evidence ^[11]. The observed spacing sits at the bottom of the modelled range ^[3][11].

The third line of evidence is chemical. Lead isotopes in both chains share a signature that takes more than a billion years to develop, at least 17 times longer than the 60 million years the nearby subducted slab has spent down there ^[13][19]. The signature came from deep mantle, not from the slab the plume ran into ^[13]. By itself it cannot exclude two deep sources of similar composition. The authors put that half of the case as a tendency, writing that two unrelated plumes from separate patches of deep mantle "should generally not produce the same rock" ^[8].

The obstacle itself turns up in a reconstruction of the last 200 million years of plate motion. Mapping where old seafloor sank beneath the Tasman Sea shows a ribbon of slab material with gaps on either side, in the places the model puts each branch ^[12].

In the simulation, the branching does not last. As the obstructing slab sinks deeper, one branch is uncovered and becomes dominant while the other shuts down ^[14]. Eruption volumes have been declining along the Lord Howe chain for the past 23 million years or so, while the youngest Tasmantid volcanoes are getting larger ^[15]. That decline spans roughly 57 percent of the interval the two chains record ^[20]. The authors expect Tasmantid to become the main conduit and Lord Howe to stop growing ^[16].

Hotspot chains are used to work out how continents have moved because a plume stays roughly fixed while the plate slides over it ^[17]. In this account the surface position of a branch depends on where the gap beside the slab is, and the slab is sinking ^[9][14].