The three-dimensional velocity and temperature fields surrounding an isolated thermal plume in a fluid with temperature-dependent viscosity are measured using Particle-Image Velocimetry and thermochromatic liquid crystals, respectively. The experimental conditions are relevant to a plume rising through the mantle. It is shown that while the velocity and the isotherm surrounding the plume can be used to visualize the plume, they do not reveal the finer details of its structure. However, by computing the Finite-Time Lyapunov Exponent fields from the velocity measurements, the material lines of the flow can be found, which clearly identify the shape of the plume head and characterize the behavior of the flow along the plume stem. It is shown that the vast majority of the material in the plume head has undergone significant stretching and originates from a wide region very low in the fluid domain, which is proposed as a contributing factor to the small-scale isotopic variability observed in ocean-island basalt regions. Lastly, the Finite-Time Lyapunov Exponent fields are used to calculate the steady state rise velocity of the thermal plume, which is found to scale linearly with the Rayleigh number, in contrast to some previous work. The possible cause and the significance of these conflicting results are discussed, and it is suggested that the scaling relationship may be affected by the temperature-dependence of the fluid viscosity in the current work.

Convection from an isolated heat source in a chamber has been previously studied numerically, experimentally and analytically. These have not covered long time spans for wide ranges of Rayleigh number Ra and Prandtl number Pr. Numerical calculations of constant viscosity convection partially fill the gap in the ranges $\mathit{Ra}= 1{0}^{3} {{\unicode{x2013}}}1{0}^{6} $ and $\mathit{Pr}= 1, 10, 100, 1000$ and $\infty $. Calculations begin with cold fluid everywhere and localized hot temperature at the centre of the bottom of a square two-dimensional chamber. For $\mathit{Ra}\gt 20\hspace{0.167em} 000$, temperature increases above the hot bottom and forms a rising plume head. The head has small internal recirculation and minor outward conduction of heat during ascent. The head approaches the top, flattens, splits and the two remnants are swept to the sidewalls and diffused away. The maximum velocity and the top centre heat flux climb to maxima during head ascent and then adjust toward constant values. Two steady cells are separated by a vertical thermal conduit. This sequence is followed for every value of $Pr$ number, although lower Pr convection lags in time. For $\mathit{Ra}\lt 20\hspace{0.167em} 000$ there is no plume head, and no streamfunction and heat flux maxima with time. For sufficiently large Ra and all values of Pr, an oscillation develops at roughly $t= 0. 2$, with the two cells alternately strengthening and weakening. This changes to a steady flow with two unequal cells that at roughly $t= 0. 5$ develops a second oscillation.

In order to link the geochemical signature of hot spot basalts to Earth's deep interior, it is first necessary to understand how plumes sample di↵erent regions of the mantle. Here, we investigate the relative amounts of deep and shallow mantle material that are entrained by an ascending plume and constrain its source region. The plumes are generated in a viscous syrup using an isolated heater for a range of Rayleigh numbers. The velocity fields are measured using stereoscopic Particle-Image Velocimetry, and the concept of the 'vortex ring bubble' is used to provide an objective definition of the plume geometry. Using this plume geometry, the plume composition can be analysed in terms of the proportion of material that has been entrained from di↵erent depths. We show that the plume composition can be well described using a simple empirical relationship, which depends only on a single parameter, the sampling coe cient, s c . High-s c plumes are composed of material which originated from very deep in the fluid domain, while low-s c plumes contain material entrained from a range of depths. The analysis is also used to show that the geometry of the plume can be described using a similarity solution, in agreement with previous studies. Finally, numerical simulations are used to vary both the Rayleigh number and viscosity contrast independently. The simulations allow us to predict the value of the sampling coe cient for mantle plumes; we find that as a plume reaches the lithosphere, 90% of its composition has been derived from the lowermost 260 750 km in the mantle, and negligible amounts are derived from the shallow half of the lower mantle. This result implies that isotope geochemistry cannot provide direct information about this un-sampled region, and that the various known geochemical reservoirs must lie in the deepest few hundred kilometres of the mantle.