Main

Dyke propagation is dependent on balances between buoyancy, source pressure and fracture strength3,4. Wilson and Head assume that kimberlite dykes are connected to the deep source and that the pressure gradient between the source and the dyke tip is governed by the release of copious carbon dioxide (CO2). Thus, assumptions are made about the volume of available magma, CO2 solubility and volatile composition, as well as about whether source pressure or buoyancy is dominant and about the behaviour of volatiles released into the crack tip. Wilson and Head state that 90% of the CO2 is exsolved at 2 GPa. However, CO2 becomes increasingly soluble as melts become more silica-deficient5; at 100 MPa, silica-poor basic melts can dissolve >1% CO2 and, with a linear solubility law, most if not all CO2 would be dissolved at 2 GPa. Furthermore, in carbonate-rich melts, most carbon is speciated as carbonate rather than molecular CO2, as indicated by magmatic calcite in hypabyssal kimberlites6. The Wilson and Head model overestimates the amount of volatiles available to act as an exsolving propellant. Water may be a major volatile in kimberlite2, but it only exsolves at low pressure.

In the model of Wilson and Head, volatiles are released from exsolving magma into the dyke tip with a very low pressure, resulting in very high pressure gradients and very high propagation speeds (tens of metres per second). However, experimental and theoretical studies4,7 show that the much larger buoyancy of released volatiles results in a fluid-filled fracture accelerating in advance of the magma-filled dyke, consistent with observations from kimberlite dykes8. The pressure in the volatile-filled fracture moving in advance of and accelerating away from the magma must be at least the lithostatic pressure plus the mantle fracture strength, so we question the very low pressures, except for a negligibly small region at the volatile-filled crack tip3,4. Wilson and Head infer a decelerating fracture system, whereas previous work9 on dyke nucleation indicates that acceleration is a consequence of the increase in length as dykes propagate and decompress.

There are difficulties reconciling the very short eruption times estimated by Wilson and Head and the geological complexity of kimberlites2 (C. R. Clement et al., unpublished results), which indicate prolonged multistage eruptions. Furthermore, constraints on volumes and magma supply rates through established dyke systems2 indicate eruption times of days to months rather than an hour. Wilson and Head estimate large adiabatic coolings, but these are not consistent with estimates of high emplacement temperatures (>400 °C to 1,100 °C) of kimberlitic pyroclastics and hypabyssal intrusions2,10,11.

The pipe-formation process proposed by Wilson and Head is unclear, but we envisage that it involves the principles of rock mechanics2,12, combined with large early overpressures and later underpressures associated with explosive flows2. The geology supports a progressive, multistage and long-lived failure of wall-rocks by a variety of failure mechanisms rather than catastrophic pipe formation2,12. The fluidization wave model of Wilson and Head is evidently a dynamic phenomenon. Fluidization is usually applied in geological systems using concepts from engineering13,14, in which gas flows continuously through unconsolidated granular materials. There is geological and experimental evidence that fluidization occurred in the waning pipe-filling stage of kimberlite eruptions2,13,14.

We agree with Wilson and Head that fast transport aids diamond preservation, but there are other important factors because kimberlites contain mixtures of perfectly shaped, broken and resorbed diamonds15, indicating diverse interaction histories with kimberlite magmas. Diamonds can be preserved within nodules, preventing reaction with kimberlite, and are released progressively during ascent by fragmentation of xenoliths, resulting in a range of interaction times15.