In a typical experiment, three separate phases of axial compaction are observed in the early stages (Figure 3). The application of confining pressure results in anywhere from 5 to 10 percent axial strain.

Axial strain from the application of sd varies from 3.2 percent at low sd to 17.7 percent at the highest stresses. Upon the application of temperature and flow, all experiments exhibit further strain of 4-7 percent, with the exception of the experiment conducted at 25°C. Although temperature is responsible for this strain, there is poor correlation between temperature or stress and the amount of strain. Scholz et al. [1995] called this phase of compaction "hydrothermal consolidation" and speculated that it was due to increased dissolution at highly stressed grain contacts followed by grain boundary sliding. Further tests using sands of different composition and shape, however, now suggest that this phenomenon is not chemically induced, but is likely caused by temperature induced annealing of the strain hardened jacket perhaps accompanied by some thermo-mechanical rearrangement of packing. However, the nature of this hydrothermal consolidation is not important in the framework of this study since all permeability measurements are made upon the completion of this strain. Axial strain during the remainder of the experiments is characterized by creep which amounts to no more than about 1.5 percent strain after four days. However, much or all of this strain may be due to bulging of the sample and may not represent real compaction of the sand.

Although the circulation of water is initiated concurrently with the application of temperature, permeability measurements do not begin until all T-induced strain has stabilized and the system has attained thermal equilibrium. This is done to ensure that mechanical compaction or changes of temperature are not responsible for any of the observed permeability changes. To be consistent, permeability measurements in all the experiments commence 1.3 x 104 seconds after flow has been initiated (Figure 3). Four days was chosen as the duration of most of the experiments because it was found that permeability changes followed a near exponential form (see below) with a time constant which caused permeability to flatten off in less than 4 days.

The results of the permeability measurements are shown on Figure 4a and 4b. In the four highest temperature experiments, significant reductions in permeability are observed (Figure 4a). The permeability reduction is approximately 50 percent at 125°C and more than an order of magnitude at 275°C. Conversely, at 25°C the permeability increases to twice the original value before leveling off to a constant value (reasons for this fundamental difference at 25°C will be discussed below). The experiment reproduced at 175°C shows that although the starting permeability values of the two experiments are different, the relative amount of permeability reduction is approximately the same. It is this relative reduction in permeability which is of greater importance. It should be noted that final permeabilities span three orders of magnitude over the range of temperatures tested. Figure 4b shows a series of experiments conducted at T= 175°C and sd of 25 to 90 MPa. Again, all the experiments exhibit considerable reductions in permeability. The paucity of permeability data at low differential stress is mainly a result of the higher permeabilities encountered. When dealing with a high permeability sand, a very small pore pressure difference must be maintained across the sample in order to induce flow that is slow enough to allow the sample and fluid to reach the regulated temperature. As a result of these small pore pressure differences, small amounts of drift in the pore pressure transducers give rise to unstable flow and, hence, unreliable permeability estimates. To circumvent this problem, transducers are adjusted manually and permeability is measured several times, rather than continuously, during an experiment. It should be noted that although few reliable data points exist for these low stress experiments, flow is in fact continuous as in all other runs.

From Figures 4a and 4b it is clear that there are distinct differences in the initial permeabilities, both at constant temperature and constant sd. This may be understood by examining a plot of initial permeability vs porosity as calculated from the starting porosity and the compaction strain (Figure 5). Such power law correlations between permeability and porosity are well documented from experimental studies [Zoback & Byerlee, 1976; Bernabé et al., 1982; David et al., 1994; Zhang et al., 1994], field studies [Bourbie & Zingzner, 1985] and models based on percolation theory [Zhu et al., 1995]. Although permeability usually varies with porosity cubed in high porosity rocks, exponents as high as 25 have been reported [David et al., 1994]. In Figure 5, the variability in the starting porosity and permeability is caused by the differing strain resulting from the initial and thermally induced compaction. Since the six experiments conducted at various sd have highly variable strains, the correlation of strain and permeability is well defined. Conversely, there is significant scatter in the data for experiments done at different temperatures. This scatter is partly a result of the poor correlation between temperature and T-induced compaction. Inherent differences in strain behavior between different experiments probably accounts for the remainder of the imprecision. Because the porosity measurements are made indirectly (from strain data), it should be noted that the porosities used in Figure 5 are only accurate to about 2-3 percent. However, the relative porosity differences between experiments are by comparison very accurate. So, even if the data were shifted slightly to the left or right on Figure 5, the exponent of 10.6 from the power relationship would not change dramatically.

Since there is variation of measured initial permeabilities resulting from mechanical compaction, permeability data has been re-normalized to the starting values for each experiment (Ko). This has been done so that permeability reduction between experiments may readily be compared. The re-normalized permeability curves are shown on figures 6a and 6b. Figure 6a clearly shows the temperature dependence of permeability reduction; the greatest changes occurring at high temperature. Figure 6b shows the normalized permeability evolution in the variable stress experiments. Stress appears to be correlated to the amount of permeability reduction at low deviatoric stresses but there is no clearly defined trend at higher stresses. The curves shown in figure 6 are theoretical fits derived by Aharanov et al. (1997) in the companion paper. These curves will also be discussed further below.

Go to next section

Go back to table of contents