About 68 percent of the seafloor classified along a set of magnetic profiles across the Gakkel Ridge is scored as amagmatic, meaning it lacks the magnetic signature of volcanic crust, against 32 percent that carries it. That is the central result of a study of the Arctic Ocean floor published in Nature Communications on 7 August 2026, and the authors describe the dominance of the non volcanic seafloor as unexpected. High, but not a record: the same paper lists 25 percent at the Mid-Cayman Spreading Center, at least 41 percent at the easternmost Southwest Indian Ridge and up to 70 percent at the Australian-Antarctic Discordance.
Mid ocean ridges are usually described as the planet’s longest volcanic system. Along most of them, magma rises into the gap between two separating plates, freezes into basalt, and builds new crust at a fairly steady rate. The Gakkel Ridge, which runs about 1,800 kilometres from the Lena Trough off Svalbard to the Laptev Sea off Siberia, appears to do that only intermittently. Across the surveyed swath, most of the classified seafloor is instead read as mantle rock brought up to fill the gap and altered to serpentinite.
The survey was flown half a century ago
The data are not new. Fei Zhou, Ingo Grevemeyer and Jérôme Dyment analysed 60 aeromagnetic profiles collected by Project Magnet in 1974 and 1975, flown at 300 metres above the sea surface with a reading every 250 metres and track lines spaced 9 to 17 kilometres apart. The profiles cross the ridge from 3 degrees west to 50 degrees east, covering crust from zero age out to roughly 35 million years old. The navigation, levelling and field corrections had already been applied to the archived data; the authors removed long wavelength trends from it and worked from there.
The vintage of the survey matters less here than the difficulty of getting any survey at all. As the paper puts it, thick sediments in the Amundsen and Nansen basins and the ice covering the Arctic Ocean “make difficult systematic geophysical and geological surveys beyond the ridge axis.” Ships can dredge rock from the ridge crest. Reaching seafloor that lies under ice and beneath as much as a kilometre of mud is another matter.
The authors cross checked the 1974 to 1975 lines against a second, independently navigated magnetic dataset collected in 1998 and 1999. Both the position and the amplitude of the anomalies agree where the two surveys cross, they report.
The magnetic difference between erupted basalt and exhumed mantle
The method turns on a contrast in how two kinds of rock hold magnetism. Basalt erupted at a spreading centre cools in the Earth’s field of the moment and locks in that polarity, so magmatically built seafloor carries the familiar barcode of reversals. Serpentinite forms where water reaches exposed mantle peridotite. The authors’ rationale is that its “uncoherent remanent magnetization” cannot record reversals cleanly, so it produces weak, subdued anomalies instead.
In the model, basalt and gabbro carry remanent magnetisations of 20 and 2 amps per metre. The serpentinite layer is given a low induced magnetisation of 0.5 amps per metre, which, in the paper’s words, “does not record magnetic reversals.” The team built forward models of each profile and adjusted two things at once, the layout of magmatic and non magmatic crust and the spacing of the magnetic chrons, until the synthetic anomaly matched the observed one. That is two sets of adjustable parameters rather than one, which loosens how uniquely any single fit can be read. Alteration is handled as a linear decay of basalt magnetisation away from the ridge axis.
The classification produced the headline ratio, and a second, more interesting result. Comparing the two flanks of the ridge on each profile, the authors sorted every interval into robust magmatism, where both flanks are volcanic, intermediate magmatism, where one flank is volcanic and its conjugate is not, and starved magmatism, where neither is.
Nothing new was sampled
No rock was dredged for this study, no seismic line was shot, and no camera went down. The paper is explicit that its constraint is indirect, describing the work as “overcoming the difficulties of systematic rock sampling in an environment where the basement is blanketed by thick sediments.” The dredged peridotite, basalt and gabbro plotted on the study’s first figure come from earlier expeditions, and that sampling sits on the ridge axis rather than in the sediment covered crust further out.
The authors raise the most obvious objection to their own classification and then argue it away. Frequent magnetic reversals can themselves produce subdued anomalies at ultraslow spreading rates, which they say could possibly lead to overestimating the abundance of amagmatic seafloor for those intervals. Against that they point to notable variations on the conjugate southern flank of one representative profile, and to a test across all the investigated profiles showing no statistical bias toward intervals of frequent reversals. The first they call suggestive of a negligible effect; the second, confirming. The evidence for that second test sits in the paper’s supplementary information, which this article has not read.
One caveat they leave standing bears directly on the headline figure. Resolution at these spreading rates is limited, and the paper states that better than one million years cannot be reached, because a single anomaly absorbs adjacent polarity intervals. It follows, in the authors’ own words, that “some short volcanic intervals, less than 1 Myr-long or within a period of frequent magnetic reversals, may not be taken into account.”
The paper also gives the ratio several different scopes, of which two matter here. The abstract says amagmatic seafloor dominates “~68% of the Eurasian Basin.” The results section, where the number is derived, says 68 percent “of the total seafloor” after the classified points along the profiles were interpolated onto a grid, with the gaps between flight lines set to a null value. Those are different denominators, and the second is the one the method supports. This piece uses the results version.
Independent corroboration exists but is sparse. Two short seismic and gravity profiles across the sparsely magmatic central zone suggest an alternation of normal and serpentinised crust that, the authors write, correlates well with the amplitude variation of the magnetic anomalies. Two crossings of a ridge about 1,800 kilometres long make a thin basis for a regional claim. The paper says as much about the seismic literature, calling the handful of existing active source surveys “clearly insufficient to depict the nature of the lithosphere and the melt supply from a regional perspective,” though it writes that as an argument for flying magnetics rather than as a caution about the result.
A rhythm in three phases
The temporal pattern is the part the discussion develops furthest. The variations shown in the study’s fourth figure, the authors write, “suggest that a short phase of robust melt emplacement, roughly spanning 2 Myr, is always followed by a phase of intermediate magmatism, ~4 Myr-long, and a phase of starved magmatism ~5 Myr-long,” after which the sequence appears to begin again.
The results section describes the same succession slightly more loosely, saying that short intervals of robust magmatism are “usually followed by” a longer interval of intermediate magmatism lasting up to ten million years. The discussion’s four million years is the phase length in the proposed sequence, and it sits inside that stated maximum, so the two figures are not in conflict. The change from “usually” to “always” is the sharper edit, and it happens inside a sentence that is itself hedged with “suggest that.”
The individual episodes are long by any human measure. The amagmatic domains run roughly 20 to 150 kilometres across and last roughly three to ten million years, while the robust magmatic domains, sorted by comparing conjugate flanks, are comparably wide at roughly 20 to 200 kilometres but shorter lived, at roughly one to five million years. A single flank can go quiet for far longer than that: on one western profile the model places an amagmatic spreading episode spanning 16 million years, broken only by a few small volcanic patches and one magmatic interval.
For a mechanism, the authors reach for the thing that makes Gakkel unusual in the first place. It has the slowest spreading rate on Earth, its full rate falling from about 13 kilometres per million years in the west to about 6 in the east, and, the authors write, it is the only mid ocean ridge with no prominent transform fault offsetting its axis. Seismic work has suggested lithosphere possibly up to 35 kilometres thick beneath the sparsely magmatic zone. Their proposal is that a lithosphere this cold and this strong does not pass melt continuously. Melt accumulates, then escapes in bursts.
The hedging on that proposal is worth reading in the original. The abstract stacks three qualifiers into one sentence: the observations “suggest that strong lithosphere presumably characterizing this ultraslow spreading ridge may prevent a continuous melt supply.” On whether active mantle upwelling drove the older bursts, the authors write that this “cannot be proven because the mantle dynamics keep changing, and the current state of the melting zone beneath the ridge axis cannot represent the past.” They then argue that the similarity between the ancient bursts and the present one makes it probable anyway, on intervals of roughly ten million years.
What more seismics would settle
Two short crossings are the only off axis seismic corroboration the paper offers for a classification drawn from 60 flight lines. More active source seismic profiles across the ridge flanks, and eventually a core through crust the magnetic model scores as amagmatic, would test the reading directly: the drill would either recover serpentinised mantle or it would not. Neither is proposed in the paper, which argues instead that magnetics is the practical way to survey a basin under ice.
Until something reaches the rock, the claim rests where it currently sits: on magnetism recorded from the air in the 1970s, checked against a second magnetic survey from the 1990s, against two short seismic and gravity crossings, and against rock dredged from the ridge crest rather than from the buried crust in question.