Microseismic Analysis of High Pressure, High Temperature Transformational Faulting Experiments Using Broadband Ultrasonic Sensors

Abstract:

One of the major unresolved questions in earth science is how earthquakes can occur under the high pressure/high temperature (HPHT) conditions > 350 km below the surface. One hypothesis is they occur as a result of transformational faulting in which a mechanical instability is generated from the transformation of metastable olivine to its high pressure polymorphs wadsleyite and/or ringwoodite. It was shown that Mg2GeO4 (isostructural with olivine) generates seismogenic fracture associated with the transformation to spinel (isostructural with ringwoodite) at ~4 GPa confining pressure and high temperature. However, in previous studies free PZT crystals were used as acoustic sensors, which have relatively narrow bandwidths and are subject to acoustic ringing that contaminates true AE signals. In this study, a series of experiments were performed under HPHT conditions on Mg2GeO4 undergoing the olivine to spinel transition using acoustic sensors spanning the frequency range 0.2-20 MHz. Experiments were performed under controlled deformation combined with synchrotron X-ray diffraction and radiography for in situ determination of stress and strain. In each experiment, ~1000-2000 AE events were collected which are shown to locate within the sample and in many cases cluster in groups based on cross-correlation techniques and double difference location methods (hypoDD). The presence of faulting is witnessed in tomographic images. By constructing probability density functions of the power spectral density we are able to determine how much acoustic energy is radiated at a given frequency and define the envelops where sensors of differing bandwidths and resonant frequencies operate. We have also analyzed the spectra of the waveform coda. By taking ratios of the spectral content of the coda of different events we are able to remove of the influence of the medium, as well as remove the effect of the sensor response function. This allows us to compare the source spectra of events directly regardless of location or moment magnitude. From these ratios we are able to accurately determine the relative sizes of events as well as determine their corner frequencies. Results of these analysis techniques will be described for experiments run using a variety of sensors with different resonant frequencies and bandwidths.