Does the whole of the Earth's core convect?

Higgins and Kennedy 1 used data on the behaviour of iron at high temperatures and pressures to infer that the top of the Earth's liquid core is stably stratified. Attempts to confirm this result both thermodynamically 2 and using seismological data 3 have been inconclusive. I present here geomagnetic results which may resolve the controversy. If a stratified region exists there will be no upwelling or downwelling of core fluid at the core–mantle boundary (CMB), so there will be no horizontal divergence of velocity v, that is ▿ H ·v = 0, where ▿ H = ▿ − r̂(r̂·▿) r̂ denotes the unit radial vector. This hypothesis can be tested directly, using geomagnetic data, at a few isolated points on the CMB, and local averages of ▿ H ·v can be examined over the rest of the surface. A statistical treatment of the results strongly suggests that ▿ H ·v = 0, which is a consequence of a stably stratified layer.