Primordial helium entrained by the hottest mantle plumes.

Helium isotopes provide an important tool for tracing early-Earth, primordial reservoirs that have survived in the planet's interior. Volcanic hotspot lavas, like those erupted at Hawaii and Iceland, can host rare, high 3He/4He isotopic ratios (up to 50 times the present atmospheric ratio, Ra) compared to the lower 3He/4He ratios identified in mid-ocean-ridge basalts that form by melting the upper mantle (about 8Ra; ref. 5). A long-standing hypothesis maintains that the high-3He/4He domain resides in the deep mantle, beneath the upper mantle sampled by mid-ocean-ridge basalts, and that buoyantly upwelling plumes from the deep mantle transport high-3He/4He material to the shallow mantle beneath plume-fed hotspots. One problem with this hypothesis is that, while some hotspots have 3He/4He values ranging from low to high, other hotspots exhibit only low 3He/4He ratios. Here we show that, among hotspots suggested to overlie mantle plumes, those with the highest maximum 3He/4He ratios have high hotspot buoyancy fluxes and overlie regions with seismic low-velocity anomalies in the upper mantle, unlike plume-fed hotspots with only low maximum 3He/4He ratios. We interpret the relationships between 3He/4He values, hotspot buoyancy flux, and upper-mantle shear wave velocity to mean that hot plumes-which exhibit seismic low-velocity anomalies at depths of 200 kilometres-are more buoyant and entrain both high-3He/4He and low-3He/4He material. In contrast, cooler, less buoyant plumes do not entrain this high-3He/4He material. This can be explained if the high-3He/4He domain is denser than low-3He/4He mantle components hosted in plumes, and if high-3He/4He material is entrained from the deep mantle only by the hottest, most buoyant plumes. Such a dense, deep-mantle high-3He/4He domain could remain isolated from the convecting mantle, which may help to explain the preservation of early Hadean (>4.5 billion years ago) geochemical anomalies in lavas sampling this reservoir.