At undersea structures called oceanic spreading centres, two tectonic plates split apart, and molten rock from volcanic activity solidifies to produce the crust of the sea floor. These spreading centres are separated into individual segments that are tens to hundreds of kilometres long. At the ends of the segments, shearing (side-by-side sliding) of the two plates occurs along plate boundaries known as oceanic transform faults. Since their discovery in the mid-1960s¹, these faults have been considered as sites where plate material is neither created nor destroyed. But in a paper in Nature, Grevemeyer et al.² report that this description is too simplistic. They show that, in a several-kilometre-wide region called the transform deformation zone, the crust generated at one spreading segment undergoes episodes of thinning and then regrowth as it drifts towards and past the adjacent segment.

Grevemeyer and colleagues’ work was enabled by international collaborations that have supported decades of seagoing expeditions to the world’s oceanic spreading centres. The authors analysed sea-floor topography around the intersections between spreading-segment ends and transform faults globally. At these intersections, young crust produced at one spreading segment (the ‘proximal’ segment) is adjacent to old crust that has been transported from its place of origin at the other segment (the ‘distal’ segment).

The authors found that the sea floor of the old crust in this transform deformation zone (TDZ) is consistently deeper than the sea floor in the fracture zone (FZ) — a scar of fractured crust where this old crust has already drifted past the proximal segment, and shearing has ceased (Fig. 1). This rapid shoaling (rise in sea-floor elevation) is opposite to the gradual sinking expected, considering that, in migrating from the TDZ to the FZ, the crust continues to age, and should therefore cool and become denser.

Grevemeyer et al. then examined systems of spreading segments that represent the global range of sea-floor-spreading rates. They discovered that the depth of the TDZ tends to be maximal when sea-floor spreading is ultraslow³ (slower than 20 millimetres per year) or slow (20–55 mm yr⁻¹), and minimal when spreading is relatively fast (55–140 mm yr⁻¹). To explore the physical causes of this relationship, the authors produced computer models that simulate deformation in the volume of crust and mantle below a transform fault joining two spreading segments.

As expected, Grevemeyer and colleagues found that the geometry of the plate boundary leads to shearing in the TDZ. However, by accurately simulating deformation of the brittle lithosphere (the crust and uppermost mantle)⁴, the authors determined that the TDZ also undergoes horizontal stretching. This stretching thins the lithosphere, causing the TDZ to deepen. The computer models predict that the amount of stretching and deepening increases with increasing difference in the crustal age across the TDZ. This difference increases with decreasing spreading rate because, at lower rates, the crust from the distal segment takes longer to reach the proximal segment. Therefore, the authors’ models explain the newly discovered global relationship between spreading rate and TDZ depth.

Grevemeyer et al. attribute the rapid shoaling between the TDZ and the FZ to magma produced beneath the spreading centre that periodically overshoots the proximal-segment end (for example, in magma-filled fractures) and reconstructs the TDZ crust. The evidence for this overshooting is seen in a series of J-shaped ridges that run parallel to the proximal segment and extend across the TDZ–FZ intersection, hooking inwards towards the TDZ (Fig. 1). Further signs of such volcanic reconstruction include circular volcanic domes and other hilly topographic features.

Such J-shaped ridges have been noted in studies of systems in which the spreading rate is intermediate⁵ or fast⁶. However, Grevemeyer et al. document the prevalence of these features at all spreading rates, and assert that the ridges have a widespread role in this new concept of transform-fault evolution. Moreover, the authors make the striking observation that the amount of shoaling between the TDZ and the FZ seems to be independent of the spreading rate. This finding suggests that the degree of volcanic reconstruction is as high at cool, magma-starved slow-spreading systems as it is at hot, magma-rich fast-spreading systems.

Grevemeyer and colleagues’ discoveries and interpretations are compelling, but also demand further investigation. For example, future studies should aim to reconcile the predicted horizontal stretching of the TDZ with the fact that earthquakes along oceanic transform faults are mainly associated with shearing⁷ rather than stretching⁸. Moreover, further modelling work should examine whether the results of the authors’ model persist under more-realistic conditions than those considered in their paper. For instance, such work could consider topography supported by dynamic stresses⁹, surface faults and state-of-the-art deformation laws derived from rock physics.

Regarding volcanic reconstruction, sea-floor seismic studies are needed to test the hypothesized crustal thickening and to reconcile discrepancies between the evidence presented by Grevemeyer and colleagues and that reported in a previous study¹⁰ — especially, evidence from gravity measurements that suggests crustal thickening does not occur at slow-spreading systems. Finally, it is unclear whether the magma that forms the J-shaped ridges originates beneath the spreading centre and propagates laterally, as proposed, or is generated locally below the TDZ–FZ intersection, rises mostly vertically and is guided into the J shape by lithospheric stresses. Resolving this issue will require some combination of modelling, detailed sea-floor observations, and sampling and chemical analysis of the lava rock of the ridges. Such work is also crucial for an improved understanding of the physics of magma generation and transport near all types of plate boundary.