BIG HEAD “My biggest fault is that the faults I was born with grow bigger each year." - Haruki Murakami


My research on fault zone structure has primarily focused on answering a few basic questions: does the zone of highly damaged rocks surrounding a fault (Fig.1) persist at depth, or can deep faults be modeled as relatively narrow and highly localized regions of slip? And if faults do become more localized at depth, what are the geometrical character and surrounding velocity structure of the localized surface? The answers to these questions have fundamental consequences and predictive power for earthquake rupture dynamics, radiation patterns, seismicity patterns, and ground motion prediction. In spite of their importance, we currently only have a few pieces of evidence to draw from in answering them. Observations by geologists are limited to only a handful of outcrops worldwide that represent faults which have been exhumed from seismogenic depths. By their very nature, these outcrops have undergone significant deformation in order to be observed at the surface today, and thus may not be representative of fault zone structure as it exists in situ.

Fig. 1: Schematics showing the general damage zone structure of fault zones at the regional (left) and local (right) scales.

Seismic waves propagating through fault zones provide us with direct evidence of fault zone structure (Fig. 2). However, traditional seismic tomography techniques suffer from three main limitations when applied to fault zones: standard body and surface waves have relatively low-resolution, travel time methods have decreasing ray sampling with grid size reduction, and detailed imaging requires enormous computing power. My research aims to overcome these obstacles by focusing on three specific goals: creating a database of fault-zone-sensitive seismic phases in addition to body and surface waves, employing adjoint waveform imaging with volumetric sensitivity kernels, and by making full use of supercomputing clusters.

Fig. 2 The trapped wave resonance signal from 50 events recorded on the BS array. The amplitude of each event has been normalized, and the traces have been time-shifted to align the maximum amplitude of each event. The pattern of vertical motion across the array is remarkably similar given that each event has a completely different epicentral distance, focal mechanism, and magnitude.
Fig. 3: Video of an outcrop of pulverized granite. I'm not that strong; this granite is only meters away from a major plate-bounding strike-slip fault (the San Jacinto fault) that can produce M>7.0 earthquakes. The intense shaking produced by these large earthquakes has shattered the rock under tensile stress, producing almost no deformation visible at the hand-sample scale but severely degrading the rock's strength.
::: Under Construction :::