Crustal accretion at mid-ocean ridges governs the creation and evolution of the oceanic lithosphere. Generally accepted models^1,2,3,4 of passive mantle upwelling and melting predict notably decreased crustal thickness at a spreading rate of less than 20 mm year⁻¹. We conducted the first, to our knowledge, high-resolution ocean-bottom seismometer (OBS) experiment at the Gakkel Ridge in the Arctic Ocean and imaged the crustal structure of the slowest-spreading ridge on the Earth. Unexpectedly, we find that crustal thickness ranges between 3.3 km and 8.9 km along the ridge axis and it increased from about 4.5 km to about 7.5 km over the past 5 Myr in an across-axis profile. The highly variable crustal thickness and relatively large average value does not align with the prediction of passive mantle upwelling models. Instead, it can be explained by a model of buoyant active mantle flow driven by thermal and compositional density changes owing to melt extraction. The influence of active versus passive upwelling is predicted to increase with decreasing spreading rate. The process of active mantle upwelling is anticipated to be primarily influenced by mantle temperature and composition. This implies that the observed variability in crustal accretion, which includes notably varied crustal thickness, is probably an inherent characteristic of ultraslow-spreading ridges.

Carbon in the asthenosphere is unevenly distributed, reflecting interactions between the depleted mantle and materials added by plumes or subducting slabs. Nevertheless, mid-ocean ridge basalts (MORB) from the Arctic Gakkel Ridge, far from any existent hotspots and subduction zones, are unusually carbon-rich. Here we show that the MORB along the Gakkel Ridge display systematically elevated Zn isotopes (δ⁶⁶Zn = 0.27‰-0.41‰) relative to typical global MORB (0.26 ± 0.03‰). Their heavy δ⁶⁶Zn ratios correlate with indicators for mantle source carbon enrichment. We attribute this signature to result from recycled surficial carbonates stored in deep upper mantle under the Arctic Ocean. These carbonate components were likely delivered by ancient oceanic subduction and subsequently entrained into the mantle upwelling feeding the Gakkel Ridge. Our results demonstrate that vestiges of past subduction can dominate carbon budgets beneath plume‑unaffected ridges, and act as a principal driver of large‑scale carbon heterogeneity in the upper mantle.