Permeability Loss in Granite under Hydrothermal Conditions
Recent models of earthquake recurrence, which predict repeat time for earthquakes on the same fault, suggest that periodic rupture may result from a process involving pressurization of fluids trapped and compressed by tectonic loading of the fault zone. That is, fluid pressure is raised, counteracting the compressional force that holds the fault in place (i.e., reducing effective normal stress on the fault), until slip occurs. In order for high fluid pressures to be maintained for the hundreds to thousands of years between large earthquakes, the permeability (the ease with which water flows through porous rock) of the fault zone must be very low. Otherwise, excess fluid pressures would simply drain away through the rock. This type of earthquake model requires that hydrothermal reactions (hot water reacting with rock) within fault zones lead to the gradual formation of mineral deposits that clog and fill the cracks and joints in the rock around the fault zone. This gradual sealing process will eventually trap water within the fault zone and weaken the fault when pressurized. Once an earthquake occurs and new fractures are formed within the fault zone rocks, the process begins anew. Accordingly, earthquake recurrence intervals may depend in part on the rate of permeability reduction of rocks under elevated temperature and pressure. Is this model reasonable for typical faults such as the San Andreas? To test this, we can measure in the laboratory how fast the permeability of rock is reduced by the interaction with fluids at high temperature and pressure, extrapolate this permeability reduction rate to the permeability range that we know will impede the flow of water (around 10-21 m2, or about one billion times smaller than the permeability of beach sand) and then see if the time required to reach this low permeability is roughly the same as the time interval between large earthquakes.
In this study, permeability measurements were conducted on both intact and fractured Westerly Granite at an effective pressure of 50 MPa and at temperatures from 150 to 500°C, simulating conditions in the seismogenic (earthquake-producing) zone. Permeability of the granite decreased with time, t, following the exponential relation:
For unfractured samples run between 250 and 500°C, the rate of permeability decrease, r, was proportional to temperature and ranged between 0.001 and 0.1 per day (i.e., around 0.4 and 40 decades/year loss of permeability). Values of r for the lower temperature experiments were similar to the 250°C runs. In contrast, pre-fractured samples, more typical of natural fault zones, showed higher rates of permeability decrease at a given temperature than the intact samples. Scanning Electron Microscope (SEM) and petrographic examination of the samples after the experiments were completed revealed textures that were not observed in the starting material, including corroded crystals and abundant mineral coatings of variable composition. These results suggest that fault rocks may seal at a rate consistent with earthquake recurrence intervals of typical fault zones.
Background and Objectives
Measurements of the permeability of rock under hydrothermal conditions are not new. For instance, much research has focused on how the heat from underground nuclear waste storage facilities will affect the surrounding rock. However, neither the rate at which permeability changes nor sample configurations that resemble fault zones have been adequately explored. Our current work focuses on the mineral seals that form in fault zones during the inter-seismic period (between earthquakes). The process is known as "fault healing".
The questions we address are:
- How does the permeability reduction of intact (i.e., unfractured) rock compare to fractured or gouge-bearing (finely-crushed rock) samples that are more representative of natural faults?
- What type of mineral dissolution or precipitation features will be produced, and are the pore fluids in chemical equilibrium with the rock?
- Are the resulting permeability seals consistent with recent models of earthquake cycles?
Figure 1. Configuration for rock samples showing (a) fractured rock, and (b) fractured rock with a layer of crushed quartz gouge.
Permeability was measured on cylindrical samples of Westerly Granite at 150 MPa confining pressure and 100 MPa pore pressure, representing a depth of around 3 kilometers in a fault zone. Several different sample configurations simulated various rock/fault arrangements:
- Intact samples representing undeformed country rock.
- Fractured samples representing fractured country rock near a fault (Figure 1a).
- A fractured sample with a finely-crushed quartz gouge interlayer representing a gouge-filled fault and associated fracture zone (Figure 1b).
Fluid flow was established through the rock by maintaining the fluid pressure on one side of the sample at 2 MPa above the other. This pressure difference was periodically reversed so that pore fluids flowed back and forth through the rock to establish chemical equilibrium. Permeability was calculated by measuring the steady-state flow rate of water and applying Darcy's Law:
where k is permeability; µ is the dynamic viscosity of water at the temperature and pressure of the experiment; Q is flow rate; A is the cross-sectional area of the sample cylinder and dP/dx is the fluid pressure gradient across the sample. Permeability has units of distance2.
Although flow through the fractured samples tends to be concentrated within the fracture initially, the partitioning of flow between the fracture and the adjacent rock is not known. Consequently, we plot the average or apparent permeability of the combined rock/fracture system.
- Permeability decreased with time for all samples after heating, reaching a steady rate of decrease after a few days for the intact samples to several weeks for the fractured samples (Figure 2). The rate of permeability decrease, r (from Equation 1), determined from the slope of the permeability trend at the end of each experiment, varied between 0.1 per day (10% loss per day) and 0.001 per day (0.1% loss per day), or roughly 0.4 to 40 decades per year loss of permeability. Values of r as a function of temperature (Figure 3) show that r is roughly proportional to temperature for intact samples. Fractured samples had up to an order of magnitude higher rate of decrease r than intact samples at the same temperature.
- Dissolution features on mineral surfaces were observed by Scanning Electron Microscope (SEM) on fractured samples at all temperatures and on all major mineral constituents. In addition, mineral deposits of smectite clays, feldspars, calcite, quartz, and pyrite were observed (Figure 4). These minerals are consistent with the mineral assemblages deposited in natural environments by hydrothermal solutions flowing through igneous rocks.
- Chemical analysis of pore fluids showed concentrations of major dissolved ions to be comparable to other active geothermal systems, indicating that the initially deionized water became equilibrated with the rock after repeated cycling under temperature and pressure.
Figure 2. Typical experiment at 250°C. The permeability of the granite decreases with time due to the clogging of fluid pathways by mineral deposits
Figure 3. Permeability loss in fractures and intact samples as a function of temperature. Fractured samples are sealed more easily because most of the fluid flow and hence mineral deposits are concentrated in the fracture.
Figure 4. Mineral deposits on fracture surfaces heated to 250°C. The two halves of the fractured sample have been separated to show deposits on one surface. (a). Honeycomb deposits of smectite clay on a potassium-feldspar crystal. (b). Calcite deposits on a plagioclase base.
Our analysis of the permeability loss in fractured and gouge-filled granite samples under hydrothermal conditions showed that permeability will decrease by around 2 to 6 orders of magnitude per year. With initial permeabilities in the 10-20 to 10-17 m2 range, permeability becomes negligible if the rate loss is extrapolated to typical inter-seismic time spans of around 100 years. Thus, mineral reactions effectively seal the fault zone, consistent with current models of earthquake recurrence in which fault seals are required to develop near-lithostatic fluid pressures, which in turn contribute to periodic fault rupture.