When aqueous CO2 flows through a soluble rock such as calcite, then over time the solid dissolves, sometimes leading to the spectacular erosion morphologies observed in underground caverns. An understanding of the dynamics of the dissolution process is important in predicting the long-term integrity of underground storage systems, and in particular those being proposed as part of the Department of Energy Carbon Sequestration Program.

In the standard model for dissolution, the fracture aperture is assumed to be constant. However, such models do not predict the development of deep penetration, necessary for Karst (cavern) formation, without a strong concentration dependence of the reaction rate. This kinetic “trigger” is well verified for calcite dissolution [1]; near saturation it is known that the dissolution rate drops by several orders of magnitude. However in other minerals, for instance Gypsum, the rate coefficient is essentially independent of concentration yet caverns develop in these formations as well. We have used numerical simulations [2] to investigate another mechanism for the formation of deep channels, namely flow focusing [3].

A planar dissolution front is unstable to small perturbations [4]; regions where the front is slightly extended obtain more flow, which serves to amplify the initial perturbation. The upper image shows a dissolution front advancing into a spatially homogeneous porous medium, at a point in time where the initial sinusoidal perturbations start to develop into distinct channels [5]. A selection mechanism then begins to operate, leading to the more rapid growth of long channels at the expense of the shorter ones, as can be seen in the second panel where there are fewer but deeper channels.

The mechanism for the flow focusing can be understood in terms of the model [5] shown in the inset figure. The longer channel (B) has the higher flow rate, so that the pressure gradient near the inlet is larger in the long channel (B) than in the short one (A), leading to a lower pressure in the longer channel. Thus under-saturated fluid, near the inlet, is drawn towards the long channels, further extending them at the expense of the short ones. Further into the fracture the situation is reversed. Near the outlet the pressure is higher in the long channel, as shown in the right-hand inset, so that fluid tends to flow from the channel into the surrounding porous matrix, leading to the well known “tip splitting” seen in the lower panel. This inhomogeneous growth is transport controlled, so the channel competition is qualitatively independent of the detailed reaction kinetics.