The measurements were carried out up to 0.5 GPa (wet samples) or 0.6 GPa (dry samples) at temperatures ranging up to 500°C using a piston cylinder apparatus and the techniques described above. In this experiment, the pressure was first increased to 0.5 GPa (wet samples) or 0.6 GPa (dry samples) at room temperature and then the temperature was increased to 500°C. To ensure a steady state travel time, we kept the P-T conditions of the experiment constant for 30 to 180 min before performing the velocity measurements. The sample recovered elastically and retained its original length, which suggests that it had not been subjected to plastic deformation during the entire experiment. The compressional wave velocities are plotted against pressure in Figure 3 and against temperature in Figure 4.
As shown in Figure 3, the compressional velocities in the dry sample increased rapidly until a pressure of approximately 0.3 GPa was reached. The increase in V p was probably attributable to the closure of microcracks (e.g., Birch 1960; Kern 1982). The compressional wave velocities also showed a linear increase at higher pressure between 0.3 and 0.6 GPa, which suggests that the effect of porosity on the measured velocities was not significant within this pressure range. By fitting velocity data using linear functions of pressure, we obtained the pressure derivative ∂V p/∂P = 0.29 km s-1 GPa-1. The pressure derivative (in the range P = 0.3 to 0.6 GPa) was then used to linearly extrapolate the velocity data, which yielded a V p value of 6.04 km s-1 under ambient conditions. This is the so-called intrinsic velocity of Kern (2011). As in previous studies (Shingai et al. 2001; Kitamura et al. 2003; Kono et al. 2004; Nishimoto et al. 2005; Ishikawa et al. 2008), linear trends of velocity as a function of pressure were observed above a threshold pressure of between 0.3 and 0.6 GPa. At lower pressures, the velocities of the samples are compromised by unclosed pore spaces filled with air.
The dry sample was heated from room temperature at 0.5 GPa to 500°C. Unfortunately, the lead wire was disconnected from the piezoelectric transducer within the high-pressure cell at 500°C and therefore V p could not be measured at 500°C. As the temperature increased to 400°C, the compressional velocities decreased from 6.19 to 5.99 km s-1 (Figure 4). By fitting the velocity data using linear functions of temperature, we obtained the temperature derivative ∂V p/∂T = -0.52 × 10-3 km s-1°C-1. Because the compressibility and thermal expansion of low quartz is less than 0.5% between ambient conditions and 0.5 GPa and 400°C (e.g., Raz et al. 2002), the changes of sample length can be considered to be insignificant for the determination of V p.
To understand the effect of water on V p, we measured the compressional wave velocities in wet samples (Figures 3 and 4). In each experiment involving a wet sample, the pressure was first increased to 0.5 GPa at room temperature (Figure 3). The compressional velocities increased rapidly up to a pressure of 0.3 GPa. This is probably attributable to the closure of air-filled pore spaces. The compressional wave velocities also increased linearly at higher pressures between 0.3 and 0.5 GPa, and these values showed a similar pressure dependence for different H2O contents, which suggests that the effect of pore air on the measured velocities is not significant within this pressure range. By fitting velocity data using linear functions of pressure, we obtained the pressure derivative ∂V p/∂P = 0.45 km s-1 GPa-1 for 0.37 wt.% H2O and 0.54 km s-1 GPa-1 for 0.98 wt.% H2O. As before, we used the pressure derivative to linearly extrapolate the velocity data (in the range P = 0.3 to 0.5 GPa) to ambient conditions, which resulted in an intrinsic V p of 5.79 km s-1 for 0.37 wt.% H2O and 5.65 km s-1 for 0.98 wt.% H2O. After pressurization to 0.5 GPa, the sample was heated to 500°C (Figure 4), which caused the compressional velocities of the quartz aggregate with 0.37 wt.% H2O to decrease from 6.02 to 5.79 km s-1. By fitting the velocity data using linear functions of temperature, we obtained the temperature derivative ∂V p/∂T = -0.49 × 10-3 km s-1°C-1 for quartz aggregate with 0.37 wt.% H2O. After repeating the experiment using quartz aggregate with 0.98 wt.% H2O, the compressional velocity decreased from 5.92 to 5.78 km/s as the temperature increased to 500°C. By fitting the velocity data using linear functions of temperature, we obtained ∂V p/∂T = -0.28 × 10-3 km s-1°C-1. Comparison of the temperature derivatives of wet samples with that of the dry sample showed that the temperature derivatives were insensitive to the addition of fluid H2O within the 1 wt.% range investigated.
The velocity drop is plotted against temperature in Figure 5. We used the following definition of the velocity drop:
The velocity drop was approximately 3% in the experiment using quartz aggregate with 0.37 wt.% H2O. In the experiment with higher H2O content (0.98 wt.% H2O), the velocity drop remained constant (4%) up to 200°C and then decreased slightly to 3% with increasing temperature.
Zhao et al. (1996) reported a low seismic velocity (-5%) and high Poisson's ratio (+6%) anomaly extending over approximately 300 km2 at the hypocentral zone of the magnitude 7.2 Kobe earthquake that occurred on January 17, 1995. They speculated that this anomaly was due to an over-pressurized, fluid-filled, fractured rock matrix that contributed to the initiation of the Kobe earthquake. The low velocity anomaly (-5%) is almost equivalent to the velocity drop (-4%) shown in our experimental results with 1 wt.% H2O. Our results also show that the addition of 0.4 to 1.0 wt.% H2O into the quartz aggregate causes a velocity drop of 3% to 4% at the pressures and temperatures of the middle crust (0.5 GPa, 25°C to 500°C), which suggests that one possible reason for low seismic velocities (-5%) is 1 wt.% H2O fluid-filled rock in the hypocentral zone. At this stage, we see our experimental data as a useful contribution towards improved quantitative understanding of the effect of H2O on the elastic wave velocities of rocks. We also hope our results will be useful for determining the geofluid distribution within the crust, thereby enhancing future interpretations of seismic tomographic data.